2c. Engeler, I., Barasz, K. (2021) - From Mix-and-Match To Head-to-Toe

From Mix-and-Match to Head-to-Toe: How
Brand Combinations Affect Observer Trust
Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

ISABELLE ENGELER
KATE BARASZ
Consumers use brands in many combinations, from mixing-and-matching multiple

brands (e.g., Nike shoes, Puma shirt, and Asics shorts) to using products primarily
or solely from one brand (e.g., Nike shoes, shirt, and shorts). This work explores
how such combinations affect observers’ trust in another consumer’s recommen-
dations. Comparing two combination types—mixed-brand combinations (where
all/most branded products are from different brands) and dominant-brand combi-
nations (where all/most branded products are from the same brand)—nine studies
establish that observers tend to have less trust in recommendations from those
who use dominant-brand combinations (studies 1A–1C). This is driven by infer-
ences about how the products were chosen: observers believe others who use
dominant-brand combinations placed relatively greater importance on the brand—
a feature that often serves as a mental shortcut for choices—and therefore infer
these consumers made quicker, less thoughtful decisions (studies 2A and 2B).
While the effect diminishes when observers hold particularly favorable attitudes to-
ward the focal brand (study 3), it can alter observers’ own downstream behaviors
(e.g., social media following intentions, information seeking, and recommendation
taking; studies 4A–4C). Together, the findings confirm that brand combinations
elicit responses distinct from single brands, offering fruitful avenues for future
research.
Keywords: choice inferences, brand combinations, decision thoughtfulness, con-

sumer trust, advocacy and influencer marketing, social influence
C onsumers use brands in many combinations—from

mixing and matching products from multiple brands,
to using products primarily or solely from one brand, and
anything in between. However, despite their ubiquity, very
little consumer research has examined the signals sent by
combinations of brands (Rahinel and Redden 2013), in-
Isabelle Engeler ([email protected]) is an assistant professor in the stead focusing on the signaling effects of single brands
marketing department at IESE Business School, Av. Pearson, 21, 08034
Barcelona, Spain. Kate Barasz ([email protected]) is an associate (Ferraro, Kirmani, and Matherly 2013; Han, Nunes, and
professor in the marketing department at Univ. Ramon Llull, ESADE, Av. Drèze 2010; Rice, Kelting, and Lutz 2012; Wernerfelt
Torre Blanca, 59, 08172 Sant Cugat, Barcelona, Spain. Both authors con- 1990)—the literature’s more traditional unit of analysis.
tributed equally to this manuscript. Please address correspondence to
Isabelle Engeler. The authors thank trnd Espa~na and bandido.run for help
Yet, as the present research shows, brand combinations are
in data collection; Alberto Rampullo, Eugenia Bajet, and Carla Morales worthy of broader consideration, with unique signaling
for their research assistance; and audiences at ACR (Dallas), EMAC effects over and above single brands. In other words, even
(Hamburg), AMA-CBSIG (Bern), IESE’s Experimental Research Group, holding constant a single focal brand, the broader brand
HBS’s NERDLab, HEC Paris, and the University of Lucerne for their
helpful suggestions. Supplementary materials are included in the web ap- combination can change what observers perceive and how
pendix accompanying the online version of this article. The data, code, they subsequently behave. Thus, this work explores one di-
and materials can be found on OSF: https://osf.io/vyfz3 mension of brand combinations—the repetition of a brand
Editor: Margaret C. Campbell within a given set—and examines its effect on observers’
trust in a target consumer’s recommendations.
Associate Editor: Kevin Lane Keller To illustrate, imagine two consumers who recently pur-
chased athletic apparel—and both of whom chose identical
Advance Access publication August 17, 2021
Nike running shoes. In addition to the shoes, Consumer A
C The Author(s) 2021. Published by Oxford University Press on behalf of Journal of Consumer Research, Inc.
V
All rights reserved. For permissions, please e-mail: [email protected] Vol. 48 2021
DOI: 10.1093/jcr/ucab041
562
ENGELER AND BARASZ 563
selected a variety of products from other brands: shorts brand combination users tend to be seen as less trustworthy
from Asics, a shirt from Puma, and a hydration pack from recommenders. We contribute to the nascent literature on
Brooks. By contrast, Consumer B purchased shorts, a shirt, others’ perceived decision processes, and specifically, the
and a hydration pack all from Nike. In other words, both ways in which observers evaluate how thoughtfully another
consumers are wearing identical Nike running shoes, but person’s decision was made (Hamilton, Vohs, and McGill
these shoes are situated among two different combinations 2014; Kupor et al., 2014; Schrift and Amar 2015). In addi-
of brands: a mixed-brand or dominant-brand combination, tion, we explore brand combinations, an understudied topic
respectively. Now imagine: if you wanted information in the consumer behavior literature (Rahinel and Redden

about running shoes, whose recommendation would you be 2013), and one that—given its distinctive signal poten-
more likely to trust? tial—merits investigation. We discuss the implications for
One possibility is that observers would trust both recom- marketers and the specific relevance for social media influ-
mendations equally. They have an equivalent amount of in- encer and brand advocacy strategies.
formation about each consumer—the brands and products
both are wearing—and both consumers purchased an iden-
MOTIVATING THE INQUIRY: BRAND
tical pair of Nikes. In this case, both consumers may be
equally suited to offer a trustworthy running shoe recom- COMBINATIONS AS A UNIT OF
mendation. Indeed, this outcome would suggest that the ANALYSIS
signals sent by single brands and brand combinations are
While the study of brand combinations has appeared in-
more or less the same—that the broader combination is
frequently in the literature (Rahinel and Redden 2013), the
immaterial.
use of brand combinations in everyday life is ubiquitous
A second possibility is that observers may be inclined to
(e.g., combinations of clothes, electronics, personal care
trust a recommendation from Consumer B—outfitted in
items, etc.). Yet, before examining the effect of such com-
Nike products—more than Consumer A. This dominant-
brand combination user may appear to be a passionate binations on observing consumers, two questions must be
brand advocate who knows a lot about Nike, which could addressed. First, does sufficient variability of combination
have positive spillover effects on how trustworthy observ- types exist (i.e., that dominant- and mixed-brand combina-
ers judge this person’s shoe recommendations to be. tions both occur in the world)? And second, do observers
However, a third possibility—and the one that we docu- actually notice combinations?
ment here—predicts the opposite: that observers may trust To answer the first, we examined a single category—
dominant-brand combination users (Consumer B) less than athletic apparel—and triangulated the prevalence of differ-
mixed-brand combination users (Consumer A). Indeed, ent brand combinations using three samples. (Methods and
even holding constant a focal product (e.g., Nike running results are in the web appendix in supplementary studies
shoes), we find that the broader brand combination can S1 and S2; we report the brief findings here.) For the first
serve as a cue by which others’ decision motives and pro- sample, we collaborated with an artificial intelligence firm
cesses are inferred and evaluated. Combinations in which that uses machine learning algorithms to identify the
there is a dominant brand are perceived to have been moti- brands that runners are seen wearing in race photos. Within
vated primarily by brand, thus giving the impression of a their image database, they identified 4,132 runners on
quick and less thoughtful choice, while mixed-brand com- whom brands were discernible for three clothing items
binations appear to have been more thoughtfully curated. (shoes, shorts/pants, and shirts). The data revealed a variety
Consequently, observers tend to find recommendations of combinations (figure 1A).
from dominant-brand combination users less trustworthy For the second sample, we examined posts on
than those from mixed-brand combination users, which Instagram, a social media platform on which people post
affects their willingness to take recommendations for both photographs and videos. Two research assistants manually
the focal item (e.g., running shoes) and other items in the assembled and coded a sample of 500 publicly visible
category (e.g., running gear). Instagram posts with the hashtag #running, on which three
Therefore, this article explores the effect of brand com- or more branded items could be detected. We excluded any
binations on observers’ trust in others’ recommendations. posts containing explicit sponsorship disclosures, resulting
We compare combinations at each end of the brand repeti- in 489 usable posts that included three branded items (352
tion spectrum: mixed-brand combinations, where all or posts), four branded items (107 posts), or five or more
most branded products are from different brands, and dom- branded items (30 posts). The results (figure 1B) revealed
inant-brand combinations, where all or most branded prod- an assortment of different combination types.
ucts are from the same brand. Across different product For the third sample, we surveyed consumers on
categories (running gear, clothing, toiletries, electronics, Amazon’s Mechanical Turk (N ¼ 506, 45% female; Mage ¼
baby gear) and brand types (real vs. fictitious brands), we 39.66, SD ¼ 11.48) about their everyday brand combina-
find converging evidence of the effect: that dominant- tion usage in activewear; of those surveyed, 476 indicated
564 JOURNAL OF CONSUMER RESEARCH
FIGURE 1
TRIANGULATION OF PREVALENCE OF REPEATED BRAND USAGE—EXAMPLE OF ACTIVEWEAR.

Note: For interpreting (A) and (B), “3 of 3” means that three out of the three branded items were all from the same brand, which is equal to “100%” or “Fully
Dominant”.
that they use activewear. To these activewear users, we the data rely on retrospective self-reports, which could suf-
asked what combination of brands they wear together on a fer from poor recall or experimenter demand (e.g., claim-
typical occasion; answers were recorded using a slider bar ing to notice to be a “good participant”), they suggest that
anchored “all/most from the same brand” (0) to “all/most people not only observe single brands (the literature’s tra-
from different brands” (100), with midpoint labeled “some ditional unit of analysis), but a sizable majority also report
from the same and some from different brands” (50). The noticing brand combinations. Importantly, this was true of
results (figure 1C) revealed that more people tend to use both dominant- and mixed-brand combinations; while peo-
mixed-brand combinations, but dominant-brand combina- ple were more likely to notice mixed-brand (vs. dominant-
tions were not infrequent. brand) combinations, both were reportedly common (i.e.,
Finally, to answer the second motivating question above 50%).
above—do observers actually notice brand combina- Together, these data motivate our investigation: People
tions?—we administered one preregistered survey to two use an array of different combinations across product cate-
large participant samples: (1) Amazon Mechanical Turk gories and report noticing the combinations that others are
(MTurk) users (N ¼ 496, 50% men; Mage ¼ 40.3, SD ¼ using and wearing. Yet, it is important to note the heteroge-
12.5; https://aspredicted.org/ig6a2.pdf), and (2) consumers neity across contexts, categories, and consumers. For in-
recruited in collaboration with an influence marketing stance, dominant-brand combination usage was more
agency, using their highly involved pool of individuals prevalent on Instagram than in self-report data, perhaps
who are regularly recruited to test and/or promote products driven by undisclosed “paid influencer” arrangements.
from contracted brand campaigns (N ¼ 4,007, 82% women; Likewise, self-reported noticing was higher among the
Mage ¼ 37.9, SD ¼ 10.6; https://aspredicted.org/ consumer sample than the MTurk sample, perhaps because
6m2qs.pdf). (See table 1 for questions and results.) While this was a particularly involved group accustomed to
TABLE 1
SURVEY STIMULI AND RESULTS

interacting with—and more “on the lookout for”—brands. and Aronson 1965; Mills and Harvey 1972), or likeability
Extrapolating, this suggests that not all people notice (Mills 1966). Other factors pertain instead to features of
brands in equal measure. Whether due to individual or situ- the recommender’s decision: what the person chose and
ational differences, not everyone pays attention to brands via what (actual or presumed) process. For example, peo-
(or brand combinations) all the time; therefore, we would ple are more willing to follow the advice of recommenders
naturally only expect people to make inferences about who appear to have made thoughtful choices (Hamilton et
brand combinations conditional upon noticing them. al. 2014; Kupor et al. 2014; Schrift and Amar 2015).
Regardless, these data help establish that brand combina- But what exactly signals that a decision was thoughtful?
tions—both dominant and mixed—are indeed a phenome- It is within this domain—how people come to infer and
non that consumers use and notice. We now explore the evaluate others’ decision processes—that we situate the
inferences that can occur as a result. present research. While the topic has (as yet) been under-
studied (Kupor et al. 2014; Schrift and Amar 2015), some
CONCEPTUAL DEVELOPMENT research has demonstrated how inferences about another
person’s decision process can change judgments about
Consumption is an inherently social endeavor. When both that target other’s thoughtfulness and one’s own be-
forming opinions or making purchase decisions, consumers havior. For instance, witnessing someone else’s decisional
attain information from any number of sources, but perhaps conflict (e.g., a pained facial expression during a choice)
none more important than their peers (Bikhchandani, leads consumers to make similar choices (Schrift and
Hirshleifer, and Welch 1992; Carl 2008; Lee and Koo Amar 2015). Relatedly, knowing that a decision-maker
2012; Ward 1974; Watts and Dodds 2007). However, not exerted effort or spent time making a choice increases per-
all peers are created equal; a number of factors can affect ceived decision thoughtfulness and recommendation fol-
the extent to which a source is seen as credible (for review, lowing (Kupor et al. 2014). Moreover, even when people
see Pornpitakpan 2004), and in turn, influential (Cialdini are unaware of the conflict, effort, or time that went into
2008). Some of these factors pertain to specific characteris- making a decision, they use available cues to make
tics of the recommender (Giffin 1967)—for example, their assumptions about the underlying process. For instance,
levels of expertise (Busch and Wilson 1976; Doney and simple linguistic markers in speech (Hamilton et al. 2014)
Cannon 1997; Moorman, Desphande, and Zaltman 1993; and deviatory sentiments in online reviews (Kupor and
Sah, Malaviya, and Thompson 2018), attractiveness (Mills Tormala 2018) tend to make recommenders appear more
FIGURE 2
CONCEPTUAL FRAMEWORK
Why was the choice made? How was the choice made?
Relative
Decision
Importance of
Thoughtfulness
Brand
B C

Dominant-
vs. Mixed-Brand Trustworthiness
Combination
A D
thoughtful, likewise encouraging consumers to follow their What are the implications of this? For this, we draw the
recommendations. link between the relative importance of brand and per-
In this vein, we suggest that brand combinations—and ceived decision thoughtfulness (nodes B to C). Arguably,
specifically, the repetition of brands within the combina- one of the primary purposes of brands is to serve as a men-
tion—can serve as cues about decision thoughtfulness. In tal shortcut or heuristic for consumers (Hoyer and Brown
particular, we focus on more homogeneous “dominant- 1990; Maheswaran, Mackie, and Chaiken 1992; Simonson
brand” combinations, in which all or most products are and Winer 1992), who use brand knowledge to bypass an
from the same brand (e.g., Consumer B from the opening exhaustive information search or extensive prepurchase
example) and more heterogeneous “mixed-brand” combi- diligence. It follows, then, that people not only use brands
nations, in which all or most products are from distinct to make faster decisions for themselves, but may also per-
brands (e.g., Consumer A). We propose that such brand ceive that others have done the same. Therefore, we posit
combinations can change people’s trust in peer recommen- that brand-driven decisions are often perceived to have
ders via a distinct two-step process: observers may use been quicker and thus less thoughtful.
brand combinations to make inferences about (1) why a Why does this matter? We close by drawing the link be-
target consumer made a given decision (figure 2, node A to tween perceived decision thoughtfulness and trust (nodes C
B), which then affects inferences about (2) how the deci- to D)—or the degree to which an observer perceives a rec-
sion was made—or decision thoughtfulness (figure 2, node ommender’s assertions to be valid (Hovland, Janis, and
B to C). Kelley 1953). Trustworthiness is one of the primary tenets
To unpack the process, we first examine the link be- of source credibility (Pornpitakpan 2004) and has been
tween brand combination and relative importance of brand shown to affect things like a consumer’s openness to being
(nodes A to B). Research has shown that when observing influenced by a peer (Kupor et al. 2014) and one’s willing-
others’ decisions, people rely upon features of a chosen op- ness to purchase a recommended product or service (Hsiao
tion to infer decision motives (Barasz, Kim, and et al. 2010). Therefore, we suggest that—because observ-
Evangelidis 2019); in particular, prominent features are as- ers tend to perceive dominant-brand combination users’
sumed to have disproportionately motivated the choice. decisions to be less thoughtful—they will also tend to find
(For example, if someone else buys a product with a partic- their recommendations to be less trustworthy, both for the
ularly prominent color, observers tend to believe that per- focal product (e.g., the Nike running shoe) and for the cate-
son cared most about—and chose the product specifically gory more broadly (e.g., running gear). This, in turn, has
because of—the color.) As applied to the present research, the potential to change observers’ own downstream
for dominant-brand combinations—in which a single brand behaviors.
is repeated across multiple products—brand is the promi-
nent feature. As such, we suggest that consumers encoun- OVERVIEW OF STUDIES
tering a dominant-brand combination (e.g., someone in
head-to-toe Nike running gear) will tend to conclude that The scope of our investigation is as follows. We exam-
brand disproportionately motivated the choice (e.g., “She ine contexts in which consumers casually encounter one
must have chosen the running shoes because of the another—perhaps in a store, or on Instagram, or through a
brand”), and was both more important than other features favorite blog—and have the opportunity to make infer-
(e.g., quality, style, fit, etc.) had been to the dominant- ences about others’ choices. Because our theory rests upon
brand combination user, and more important than brand the inferences these observers make about others’ purchase
had been to a mixed-brand combination user. decisions (figure 2, nodes B and C), it follows that
observers must believe a purchase decision was made; Method

therefore, we do not investigate professional endorsements
Participants. Participants were recruited from
or sponsorships in which products were known to be gifted
Amazon’s Mechanical Turk (N ¼ 200; 58% male; Mage ¼
or chosen without volition (e.g., a sponsored athlete who is 35.72 years, SD ¼ 14.49).
known to be contractually obligated to wear one brand).
We compare observers’ inferences across different types Procedure. Study 1A used a two-condition, between-
of brand combinations. We hold the combination size (e.g., subjects design. All participants saw the same Instagram
four running gear items, six toiletry items) constant across photograph of a runner. They were asked to imagine en-

conditions and vary only a focal brand’s multiplicity—the countering the post and tapping on the photo to see what
number of times it repeats within a given combination; brands that person was wearing. Depending on the condi-
higher multiplicity yields more homogeneous dominant- tion, participants either saw four Nike-branded products
brand combinations. As in the opening example, we test (dominant-brand combination condition) or four products
several scenarios in which the focal brand appears on every from different brands (mixed-brand combination condition;
item in a combination; however, we also find that the effect see figure 3). The Nike running shoes served as our focal
item, which was identical between conditions.
persists when the combination is predominantly—but not
As the dependent measures, participants reported their
exclusively—comprised of a focal brand (e.g., the focal
trust in the target consumer’s specific product recommen-
brand appears on five or six out of seven products; study
dations (“How much would you trust this consumer for rec-
1B). As such, “dominant-brand combinations” are defined
ommendations about shoes?”) and general category
as being predominantly, but not necessarily exclusively, recommendations (“How much would you trust this con-
comprised of one focal brand. We examine the effect of sumer for running gear recommendations?”). These
such combinations on trust in a recommender—that is, the “specific trust” and “general trust” measures were recorded
extent to which people are open to following the recom- on a slider bar anchored at “Not at all” and “Extremely”;
mender’s advice (Kupor et al. 2014). for this and all future slider bar measures, numerical values
The article is structured as follows. The first set of stud- were hidden, but answers were recorded on a 0–100 scale.
ies (studies 1A–C) examined the main effect: that observ- To ensure the stimuli were seen as naturalistic, we asked
ers tend to trust dominant-brand combination users’ participants how realistic they think the post is (1 ¼ “Not
recommendations less. The second set of studies (studies at all” to 7 ¼ “Extremely”).
2A and B) investigated the mechanism by measuring and This and all subsequent experiments concluded with de-
manipulating our proposed serial mediators. Study 3 exam- mographic questions. No data were excluded and we report
ined a boundary condition, and studies 4A through C all conditions and measures. For all Mechanical Turk stud-
assessed the downstream consequences on observers’ own ies, the sample sizes were preset with the objective of col-
behavior. lecting at least 100 participants per cell; for studies where
participants were recruited elsewhere, data collection was
set to end at a previously specified time (reported in the in-
dividual studies below) with the goal of collecting as much
STUDIES 1A–1C: THE EFFECT OF BRAND data as possible within a set timeframe.
COMBINATIONS ON TRUST
Results and Discussion
Our first three studies examined the main effect: How do
brand combinations (dominant vs. mixed) affect observers’ As intended, the realism of the Instagram post did not vary
trust in a target consumer’s recommendations? All three by condition (MDominant ¼ 4.81, SD ¼ 1.69 vs. MMixed ¼
studies used branded clothing items and measured observ- 5.14, SD ¼ 1.55; t(198) ¼ 1.44, p ¼ .152, d ¼ 0.20).
ers’ trust in the target consumer’s recommendations. Despite wearing identical Nike running shoes, partici-
pants reported less trust in shoe recommendations from the
dominant-brand combination user (M ¼ 46.66, SD ¼
29.77) than from the mixed-brand combination user
STUDY 1A: INSTAGRAM POSTS (M ¼ 56.92, SD ¼ 27.80; t(198) ¼ 2.52, p ¼ .013, d ¼
0.36). This also held for more general recommendations
Study 1A used mock Instagram photographs to examine in the running gear category: dominant-brand combination
inferences in a naturalistic setting. We leveraged the fact users were less trusted (M ¼ 52.10, SD ¼ 31.11) than
that users often tag specific brands in their posts, which are mixed-brand combination users (M ¼ 62.59, SD ¼ 25.75;
then revealed when other users “tap” on the photo—thus t(198) ¼ 2.59, p ¼ .010, d ¼ 0.37).
creating stimuli that mimic one way in which brand combi- To ensure that the dominant-brand combination user was
nations may be encountered in everyday life. not less trusted because they were wearing a less well-
FIGURE 3 seen as atypical and thus untrustworthy—by directly mea-

suring (and then controlling for) perceived typicality.
INSTAGRAM STIMULI (STUDY 1A)
Method
38.80 years, SD ¼ 11.95).

Procedure. All participants were told to imagine see-
ing a person wearing an array of items and brands on the
street. To depict the array, participants were shown a table
with two columns: one that listed seven clothing items
(work bag, wallet, scarf, shirt, shoes, belt, and jacket) and
one that listed the brand of each item. Across seven condi-
tions, we varied the multiplicity of the focal brand (Gap)—
that is, the number of times it repeated within the combina-
tion from one to seven, inclusive. When the multiplicity
was one, the focal brand appeared only once (i.e., Gap
work bag, J. Crew wallet, Express scarf, Levi’s shirt, Zara
shoes, Benetton belt, and H&M jacket); when the multi-
plicity was two, the focal brand appeared twice (the work
Note: Dominant-brand combination on left, mixed-brand combination on right. bag and wallet were from Gap); when the multiplicity was
Focal item: Nike shoes. three, the focal brand appeared three times (the work bag,
wallet, and scarf were from Gap), and so on. As such, the
regarded brand, we also tested the stimuli to measure “multiplicity of one” condition served as a conceptual rep-
how favorably or unfavorably (1 ¼ “Not at all favorably” lication of study 1A’s mixed-brand combination condition
(i.e., no repeating brands), while the “multiplicity of sev-
to 7 ¼ “Extremely favorably”) participants viewed the fo-
en” condition served as the conceptual replication of study
cal brand (Nike) and each of the other brands that were dis-
1A’s dominant-brand combination condition; the other five
played (Brooks, Puma, and Asics). A separate group of
conditions tested configurations in between.
participants (N ¼ 100; 32% female; Mage ¼ 36.6, SD ¼
We then asked participants how much they would trust
11.5) rated Nike favorably in both the dominant-brand
this consumer for recommendations about work bags
combination (M ¼ 5.43, SD ¼ 1.39) and mixed-brand com-
(“Specific trust”). Additionally, we asked them how likely
bination (M ¼ 5.14, SD ¼ 1.46; t(98) ¼ 1.01, p ¼ .313)
they believed it would be to encounter someone dressed
conditions—and significantly more favorably than the
like this on the street (“Perceived typicality”). Both were
other three brands (both conditions: MBrooks ¼ 4.33, SD ¼
measured on a 0–100 slider bar anchored at “Not at all”
1.26; MPuma ¼ 4.90, SD ¼ 1.28; MAsics ¼ 4.46, SD ¼ 1.16;
and “Extremely.”
all ps < .02), suggesting that differences in trustworthiness
were not driven by the dominant-brand combination’s re-
peated use of a less desirable brand.
A one-way ANOVA revealed a significant difference in
STUDY 1B: VARIOUS COMBINATIONS trust across the seven conditions (F(6, 693) ¼ 6.72, p <
.001, gp2 ¼ .05). To probe the results, we conducted t-tests
As our account suggests, it need not be the case that to compare all conditions against the “multiplicity of one”
dominant-brand combinations contain exclusively one condition (i.e., the fully mixed-brand combination condi-
brand. Instead, as specified above, dominant-brand combi- tion); full results are reported in table 2. Again, participants
nations are those in which a brand appears multiple times, were less likely to trust the fully dominant-brand combina-
such that it is perceived to be a dominant, focal element of tion user (“multiplicity of seven” condition; M ¼ 38.80,
the choice. Therefore, study 1B manipulated the multiplic- SD ¼ 28.27) than the fully mixed-brand combination user
ity of the focal brand—for example, having a brand appear (“multiplicity of one” condition; M ¼ 56.62, SD ¼ 24.72;
on five, six, or seven out of seven products—to empirically t(199) ¼ 4.75, p < .001, d ¼ 0.67). This significant dif-
confirm this definition. ference also held for the “multiplicity of five” and
In addition, study 1B further addressed an alternative ex- “multiplicity of six” conditions (table 2). Critically, these
planation—that dominant-brand combination users are three comparisons held in an ANCOVA controlling for
TABLE 2 combination is mostly comprised of one repeated brand,

MULTIPLICITY OF FOCAL BRAND (STUDY 1B) such that the brand is perceived to be the dominant element
of the combination. (This is discussed further in the
Trust in recommendation Typicality General Discussion section.) Second, the results confirm
Multiplicity that reduced trust in dominant-brand combination users
M p d M p d
cannot be explained by atypicality. (We conducted typical-
1 [Fully mixed] 56.62 [Reference] 56.03 [Reference] ity pretests for all stimuli, across all studies, which confirm
2 53.93 .459 0.11 57.44 .723 0.05
3 55.18 .678 0.06 62.14 .137 0.21 this notion; see web appendix.) In fact, while observers

4 55.94 .837 0.03 64.96 .021 0.33 trusted consumers with five or six items of the same brand
5 48.39 .015 0.35 63.19 .062 0.27 less than fully mixed-brand combination users, they judged
6 48.35 .018 0.34 57.06 .788 0.04
7 [Fully dominant] 38.80 <.001 0.67 38.62 <.001 0.60 these combinations to be just as typical as fully mixed-
brand combinations.
Note: Table should be read as means compared against the “multiplicity of
one” condition as reference. For example, “multiplicity of five” compares Having established this, the studies that follow compare
mean trust from this condition (48.39) against the mean trust from “multiplicity combinations at either end of the repetition spectrum—
of one” (56.62; p ¼ .015, d ¼ 0.35). combinations that are “fully dominant” (i.e., all of the
same repeated brand) and combinations that are “fully
FIGURE 4 mixed” (i.e., no repeated brand)—in an effort to cleanly
test the phenomenon and its mechanism.
TRUST IN RECOMMENDATIONS BY CONDITION,
CONTROLLING FOR PERCEIVED TYPICALITY (STUDY 1B)
STUDY 1C: DIRECTION OF THE EFFECT
To rule out the possibility that specific brands drove our
effect, study 1C tested the phenomenon using fictitious
clothing brands (e.g., Brand A, Brand B, Brand C, etc.). It
also included a control condition to investigate the direc-
tion of the effect: Relative to baseline, do people trust
dominant-brand combination users less (as our account
predicts) or mixed-brand combination users more?
Method
Note: White bar represents comparison condition (“multiplicity of one”); light 34.93 years, SD ¼ 12.37).
gray bars are not significantly different from comparison condition, while dark
gray bars are significantly different (ps < .009) from comparison condition. Procedure. Study 1C asked participants to imagine
Error bars represent 95% confidence intervals. seeing a person on the street who was wearing five articles
of clothing. Participants saw a table with two columns: one
typicality (ps < .009; see figure 4), suggesting that atypi- that listed five clothing items (jacket, belt, shoes, work
cality cannot explain the results. bag, and wallet) and one that listed the brand of each item.
The three comparisons also held (all p < .019) when ad- The items were either labeled as being from Brand A,
ditionally controlling for typicality and the brand combina- Brand B, Brand C, Brand D, and Brand E (mixed-brand
tion typicality interaction term. This latter ANCOVA combination condition) or all from Brand A (dominant-
yielded a main effect of typicality (F(1, 686) ¼ 83.87, p < brand combination condition). In a third condition (con-
.001, gp2 ¼ .11) and brand combinations (F(6, 686) ¼ trol), participants only received information about one
2.94, p ¼ .008, gp2 ¼ .03) but only a marginal interaction branded item (rather than five): the work bag from Brand
(F(6, 686) ¼ 2.05, p ¼ .057, gp2 ¼ .02), implying that the A. Thus, the focal item in all three conditions was the work
more atypical the person is perceived by an observer, the bag from Brand A. (Exact stimuli in web appendix.)
lower their trust—but that only marginally varied across As in study 1A, participants used a slider bar to specify
the brand combination conditions. (1) how much they would trust this consumer for recom-
The results empirically confirm our definition of mendations about work bags (“Specific trust”) and (2) how
dominant-brand combinations. Dominant-brand combina- much they would trust this consumer for clothing recom-
tions need not, strictly speaking, contain exclusively one re- mendations (“General trust”) (0 ¼ “Not at all” to 100 ¼
peated brand; observers’ trust may also be diluted when a “Extremely”).
FIGURE 5 brand combination users. This occurred using Instagram

stimuli tagged with real brands (1A), different configura-
TRUST IN RECOMMENDATIONS (STUDY 1C) tions of dominant-brand combinations (1B), and fictitious
brands (1C).
For converging evidence and to rule out alternative
Control explanations, we ran three additional studies included in
70 Mixed-brand the web appendix. Supplementary study S4 showed that
Dominant-brand the effect also persists in a within-subjects design and for
Trust in Recommendaon

60 both luxury and nonluxury products. Supplementary study
S5 demonstrated that the effect holds for both experts and
novices: experts were seen as more trustworthy than novi-
50
ces, but dominant-brand combination experts still tended to
be less trusted than mixed-brand combination experts.
40 Supplementary study S6 tested the effect using different
combination sizes of two, three, four, five, and six items;
with the exception of a two-product combination (p ¼
30 .102, d ¼ 0.23 for general trust and p ¼ .337, d ¼ 0.14
Specific Trust General Trust for specific trust), the effect is robust to different combina-
Note: Error bars indicate 95% confidence intervals. tion sizes (all p-values between .018 and <.001). Thus,
having manipulated many factors (i.e., product category,
Results stimuli type, brand type, experimental design, product sta-
tus, expertise, combination composition, and combination
A one-way ANOVA again revealed a significant main size), we suggest that the effect is not merely an artifact of
effect on both specific (F(2, 297) ¼ 12.99, p < .001, gp2 ¼ the specific stimuli we selected, but is more broadly gener-
.08) and general trust (F(2, 297) ¼ 15.80, p < .001, gp2 ¼ alizable. In the studies that follow, we examine why the ef-
.10). Examining the dominant- versus mixed-brand combi- fect occurs.
nation conditions, pairwise comparisons revealed that par-
ticipants trusted the dominant-brand combination user
STUDIES 2A–2B: THE MECHANISM
significantly less for specific recommendations about work
bags (MDominant ¼ 40.83, SD ¼ 28.57 vs. MMixed ¼ 56.36, UNDERLYING THE EFFECT
SD ¼ 21.35; t(198) ¼ 4.35, p < .001, d ¼ 0.62) and We suggest that observers may use brand combinations
also for general recommendations about the product cate- as cues about others’ decision processes—even holding
gory (MDominant ¼ 38.78, SD ¼ 29.14 vs. MMixed ¼ 56.76, constant (or remaining altogether agnostic about) the actual
SD ¼ 21.09; t(198) ¼ 4.99, p < .001, d ¼ 0.71). person consuming the brands. As we argue, dominant-
Importantly, pairwise comparisons elucidated the direc- brand combinations may seem more quickly and heuristi-
tion of the effect (figure 5): relative to the control condi- cally chosen and hence appear less thoughtful than mixed-
tion, participants trusted the recommendations of brand combinations.
dominant-brand combination users less for both specific To illustrate, imagine trying to guess what motivated a
trust (MDominant ¼ 40.83, SD ¼ 28.57 vs. MControl ¼ 55.62, mixed-brand combination user’s choice of toiletry prod-
SD ¼ 22.48; t(199) ¼ 4.08, p < .001, d ¼ 0.58) ucts. As no single element within the combination stands
and general trust (MDominant ¼ 38.78, SD ¼ 29.14 vs. out, it may not be particularly evident which feature (e.g.,
MControl ¼ 53.12, SD ¼ 20.56; t(199) ¼ 4.03, p < .001, brand, quality, scent, packaging) the mixed-brand combi-
d ¼ 0.57). Said differently, the effect was not driven by nation user cared most about. By contrast, one may be
participants’ trusting mixed-brand combination users more more inclined to speculate about what motivated a
than the control (Specific trust: MMixed ¼ 56.36, SD ¼ dominant-brand combination user; their combinations have
21.35 vs. MControl ¼ 55.62, SD ¼ 22.48; t(197) ¼ .24, p ¼ a prominent, standout feature—the repeated brand—which
.811, d ¼ 0.03; General trust: MMixed ¼ 56.76, SD ¼ 21.09 observers are likely to infer was the most important feature
vs. MControl ¼ 53.12, SD ¼ 20.56; t(197) ¼ 1.23, p ¼ .219, driving the choice (Barasz et al. 2019). As such, we suggest
d ¼ 0.17).
observers will tend to infer that dominant-brand combina-
tion users made a brand-driven choice—that is, that they
Studies 1A–1C Discussion used brand as a heuristic for their decision—and therefore
Taken together, studies 1A–1C showed that participants must have been less thoughtful when deciding which item
trusted dominant-brand combination users less than mixed- to buy. This, in turn, reduces observer trust.
To confirm this account, we tested our two inference- participants how important they believed two features—
based serial mediators (figure 2, nodes B and C): (1) the brand and quality—were to the decision-maker, and then
relative importance of brand and (2) decision thoughtful- used the difference as a measure of the relative importance
ness. In study 2A, we measured these two inferences and of brand. For the latter, we asked participants how much
used mediation; in study 2B, we directly manipulated the time—a critical element of perceived decision thoughtful-
first (i.e., relative importance of brand) and measured its ness (Evans and van de Calseyde 2017; Kupor et al.
effect on the second (i.e., decision thoughtfulness) to estab- 2014)—they believed the decision-maker had spent mak-
lish a causal relationship (Spencer, Zanna, and Fong 2005). ing the choice.

To rule out one potential alternative explanation—that
PRETEST: SPONTANEOUS INFERENCES mixed-brand purchases require more shopping effort,
which itself could increase trustworthiness—we specified
Before testing observers’ inferences using structured that the products had all been purchased from the same
measures, we wanted to understand what people infer store.
spontaneously: Do observers naturally—and without
prompting—make inferences about the brand (our first me- Method
diator)? And does this vary systematically by brand combi- Participants. Participants were recruited from
nation? To test this, we ran a pretest using the exact stimuli Amazon’s Mechanical Turk (N ¼ 200; 57% male; Mage ¼
from study 2A (figure 6). We asked participants (N ¼ 100; 37.09 years, SD ¼ 11.54).
50% male; Mage ¼ 36.30 years, SD ¼ 11.97) to imagine a
consumer who had purchased six toiletry items, visually Procedure. As in the pretest, we asked participants to
depicted, which were either from six different brands or imagine a consumer named Pat had made the following
one single brand. As the dependent measure, we asked par- purchases at the grocery store and showed them six toiletry
ticipants to think for a moment about what they would infer products from Dove (dominant-brand combination condi-
about the consumer and his/her choices and write their tion) or from six different brands (mixed-brand combina-
thoughts in an open-ended text box. tion condition). The focal item held constant in both
A research assistant (blind to the conditions) coded the conditions was the Dove body lotion (figure 6).
data for two factors: (1) “inferences about brand”—if the To assess the relative importance of brand, we asked
participant mentioned the word brand or any of the specific participants two questions (counterbalanced): (1) how
brand names that were displayed (e.g., Dove, Aveeno) and much they believed the target consumer’s choice of body
(2) “inferences about product type,” which served as a con- lotion was based on the brand (0 ¼ “Not at all” to 100 ¼
trol (neutral inferences)—if the participant mentioned the “Extremely”) and (2) how much they believed the target
words skincare, body care, or any of the specific product consumer’s choice of body lotion was based on the product
types that were displayed (e.g., shampoo, body wash). We quality (0 ¼ “Not at all” to 100 ¼ “Extremely”). (Note that
compared the incidence of these two factors between and any second attribute could have been used to calculate rela-
within the two conditions. tive importance—e.g., style, packaging, scent, etc.; we
A vast majority of people (81.6%) in the dominant- chose “quality” because it applies broadly across product
brand combination condition made spontaneous mention of types and stimuli.) To assess perceived thoughtfulness, we
brand, while virtually no one did (8.2%) in the mixed- asked participants how much time they believed the target
brand combination condition (v2(1, N ¼ 100) ¼ 53.45, p < consumer had spent thinking about which body lotion to
.001, Cramer’s v ¼ .74). By comparison, most people men- buy (0 ¼ “Not at all” to 100 ¼ “A lot”). Finally, as the key
tioned product type in the mixed-brand combination condi- dependent measure, we measured participants’ specific
trust in the target consumer’s recommendation for the focal
tion (83.7%), while significantly fewer did in the
item, body lotion (0 ¼ “Not at all” to 100 ¼ “Extremely”).
dominant-brand combination condition (40.8%; v2(1,
N ¼ 100) ¼ 19.15, p < .001, Cramer’s v ¼ .44). Therefore,
we suggest that—even without prompting—people do ap- Results
pear to make inferences about brand when they encounter Analyzing the dependent measure, participants were sig-
dominant-brand combinations. In other words, the link be- nificantly less likely to trust recommendations from
tween the independent variable and the first mediator dominant-brand combination users (M ¼ 50.38, SD ¼
appears ecologically valid. 26.46) than mixed-brand combination users (M ¼ 65.26,
SD ¼ 27.06; t(198) ¼ 3.93, p < .001, d ¼ 0.56).
STUDY 2A: MEASURING INFERENCES To examine the first process measure (relative impor-
tance of brand; figure 2, node B), we calculated the differ-
Study 2A measured participants’ inferences about why ence between the importance of brand and the importance
and how the decision was made. For the former, we asked of quality, where higher positive values indicate greater
FIGURE 6
BRAND COMBINATIONS DISPLAYED IN PRETEST AND STUDY 2A

Note: Dominant-brand combination in top panel, mixed-brand combination in bottom panel. Focal item: Dove body lotion.
relative importance of brand. As predicted, participants be- mediators (95% CI [22.19, 10.28]). Partitioning this to-
lieved brand was relatively more important to dominant- tal indirect effect showed that each mediator individually
brand combination users (Mbrand — Mquality ¼ 90.13 – (M1: 95% CI [12.70, 2.83]; M2: 95% CI [11.47,
67.31 ¼ 22.83) than to mixed-brand combination users 2.98]), as well as their joint serial effect (M1þ2: 95% CI
(Mbrand — Mquality ¼ 58.86 – 71.38 ¼ 12.52; t(198) ¼ [3.75, 0.41]), each significantly contribute to explain-
8.64, p < .001, d ¼ 1.22; see figure 7). ing the variation of the indirect effect of brand combina-
For the second process measure (decision thoughtful- tions on trustworthiness (figure 8).
ness; figure 2, node C), participants inferred dominant- One limitation of this approach is that the selection of
brand combination users had been less thoughtful in decid- the serial mediation model was based on conceptual ratio-
ing which body lotion to buy (M ¼ 33.64, SD ¼ 26.10) nale (i.e., we did not conduct statistical modeling to deter-
than mixed-brand combination users (M ¼ 58.68, SD ¼ mine best fit). Furthermore, as is true for any mediation
25.71; t(198) ¼ 6.83, p < .001, d ¼ 0.97). analysis, the link between M1 (relative importance of
Examining these two measures together, a mediation brand) and M2 (decision thoughtfulness) is only correla-
analysis (95% CIs with 5,000 bootstrap samples; model 6, tional. Therefore, to further validate the model and estab-
Hayes 2018) entering the two inferences—relative impor- lish a causal link between the two mediators, study 2B
tance of brand (M1) and decision thoughtfulness (M2)— directly manipulated the relative importance of brand to
confirmed a significant total indirect effect of brand combi- demonstrate the downstream effect on perceived
nation on trust in recommendation via the two serial thoughtfulness.
FIGURE 7 (dominant-brand combination) or all from different brands

(mixed-brand combination). Dove body lotion was the fo-
RELATIVE IMPORTANCE OF BRAND (STUDY 2A) cal product.
The second factor (relative importance of brand) manip-
ulated what participants knew about how relatively impor-
100 Brand tant brand was to the target consumer. Half of participants
Quality saw only the above stimuli (thus replicating the design of
study 2A). The other half of participants also learned: “On
Importance Rang
80

a survey about his decision, Pat indicated that he cared
about both brand and quality when purchasing these
60 items.” Beneath that, participants saw a screenshot of two
survey questions Pat had supposedly filled out; they were
the exact questions we used in study 2A to measure the
40 first mediator (which was now manipulated): (1) “How
much was your choice to purchase these items based on the
20 brand?” and (2) “How much was your choice to purchase
Mixed-brand Dominant-brand these items based on the quality?” In the screenshot, the
slider bar responses were set to the exact means observed
Note: Error bars indicate 95% confidence intervals. in study 2A’s mixed-brand combination condition: 59 for
importance of brand and 71 for importance of quality (see
STUDY 2B: MANIPULATING web appendix for exact stimuli).
Using the measures from study 2A, participants reported
INFERENCES trust in the target’s recommendations and inferred decision
To provide further proof of process, we wanted to thoughtfulness (counterbalanced), which served as the two
change perceived decision thoughtfulness (M2) by manipu- dependent measures. We also measured relative impor-
lating (rather than measuring) the relative importance of tance of brand (brand vs. quality, counterbalanced), which
brand (M1). Therefore, we varied whether participants did served as a manipulation check for the explicit conditions
(or did not) receive explicit information about how rela- and was administered at the end of the study.
tively important brand was to the decision-maker.
As before, half of the participants made spontaneous Results
inferences about the target; however, we told the other half The manipulation had the intended effect of changing
that the target consumer (regardless of brand combination) the perceived relative importance of brand. A 2 (brand
had chosen the items based on both brand and quality—es- combination: dominant vs. mixed) 2 (relative impor-
sentially making explicit what participants (on average) tance of brand: inferred vs. explicit) ANOVA revealed the
had already inferred in the mixed-brand combination con- critical interaction (F(1, 598) ¼ 55.79, p < .001, gp2 ¼
dition. As such, we expected that inferences about mixed- .09): dominant-brand combination participants rated brand
brand combination users should not be affected by this ex- as relatively less important in the explicit condition (Mbrand
plicit information, but that the explicit information should – Mquality ¼ 70.58 – 71.77 ¼ 1.19) than the
attenuate the implicit inferences people make about inferred condition (Mbrand — Mquality ¼ 87.52 –
dominant-brand combination users—making them seem 55.24 ¼ 32.28; t(299) ¼ 11.22, p < .001, d ¼ 1.29). Also as
more thoughtful and, in turn, more trustworthy to intended, the manipulation did not alter inferences for
observers. mixed-brand combination participants (inferred: Mbrand –
Mquality ¼ 61.23 – 64.06 ¼ 2.83 vs. explicit: Mbrand —
Method Mquality ¼ 65.72 – 72.09 ¼ 6.36; t(299) ¼ 1.32, p ¼ .188,
d ¼ 0.15).
As preregistered, our primary analysis was a 2 (brand
combination) 2 (relative importance of brand) ANOVA
38.56 years, SD ¼ 11.80). This study was preregistered:
on decision thoughtfulness. The analysis revealed a main
https://aspredicted.org/wr66n.pdf.
effect of brand combination (F(1, 598) ¼ 42.48, p < .001,
Procedure. Study 2B used a 2 (brand combination: gp2 ¼ .07) and a main effect of relative importance of
dominant vs. mixed) by 2 (relative importance of brand: in- brand (F(1, 598) ¼ 20.87, p < .001, gp2 ¼ .03), which
ferred vs. explicit) between-subjects design. For the first were qualified by a significant interaction (F(1, 598) ¼
factor (brand combination), the stimuli were identical to 15.59, p < .001, gp2 ¼ .03). Replicating study 2A, when
study 2A: six toiletry products, either all from Dove participants made their own inferences about the relative
FIGURE 8
MEDIATION ANALYSIS (STUDY 2A)
d21 = -.16*
F(3, 196) = 21.93***
R2 = .25 Relative Decision
Importance of Thoughtfulness

Brand (M1) (M2)
b1 = -.22*** b2 = .34***
a1 = 35.35*** a2 = -19.52***
Dominant- c’ = 1.42
vs. Mixed-Brand Trustworthiness
Combination (c = -14.89***)
Total indirect effect:

-16.30, 95% CI [-22.19, -10.28]
Partitioning of total indirect effect:

M1: -7.75, 95% CI [-12.70, -2.83]
M2: -6.67, 95% CI [-11.47, -2.98]
M1+2: -1.89, 95% CI [-3.75, -.41]
Note: Figure shows the unstandardized regression coefficients and bootstrap 95% confidence intervals of the indirect effect of mediator 1 (M1), mediator 2 (M2), and
both mediators serially (M1þ2). Dominant-brand combination was coded 1 and mixed-brand combination was coded 0. ***p < .001, *p < .05.
importance of brand, decision thoughtfulness was lower To examine the two moderations in parallel (Edwards
for dominant-brand combination users (M ¼ 40.05, SD ¼ and Lambert 2007), we used a moderated mediation model
28.88) than mixed-brand combination users (M ¼ 61.54, (model 8; Hayes 2018, see figure 10). For the inferred condi-
SD ¼ 25.84; t(302) ¼ 6.84, p < .001, d ¼ 0.79); how- tions, the 95% bootstrap confidence interval showed that the
ever, when relative importance of brand was manipulated indirect effect of brand combinations on trust is mediated by
in the explicit conditions, this difference was mitigated decision thoughtfulness (Effect ¼ 13.04, 95% CI [17.10,
(MDominant ¼ 57.54, SD ¼ 24.43 vs. MMixed ¼ 62.81, SD ¼ 9.04]), but this indirect effect was significantly attenuated
20.83; t(296) ¼ 2.01, p ¼ .046, d ¼ 0.23; see figure 9, in the explicit conditions (Effect ¼ 3.20, 95% CI [6.38,
left panel). This supports a causal link between the first 0.05]). As predicted, this attenuation represents a significant
and second mediators: when brand is seen as less impor- moderated mediation effect (Index ¼ 9.84, 95% CI [4.84,
tant, decisions are seen as more thoughtfully made. 15.08]), providing further evidence for our proposed
A 2 (brand combination) 2 (relative importance of mechanism.
brand) ANOVA on trust in recommendation revealed the
same pattern: a significant main effect of brand combina- Studies 2A–2B Discussion
tions (F(1, 598) ¼ 20.74, p < .001, gp2 ¼ .03) and of rela-
tive importance of brand (F(1, 598) ¼ 15.97, p < .001, gp2 Studies 2A and 2B offer evidence of the underlying
¼ .03), qualified by a significant interaction (F(1, 598) ¼ process: When observers encountered a dominant-brand
7.38, p ¼ .007, gp2 ¼ .01). Again replicating study 2A, combination user, they inferred that the purchase deci-
when participants inferred the relative importance of brand, sion was motivated disproportionately by brand, which
trust in recommendations was lower for dominant-brand led to a belief that the choice was less thoughtfully
combination users (M ¼ 47.22, SD ¼ 26.84) versus mixed- made. As a result, observers reported lower trust in rec-
brand combination users (M ¼ 61.39, SD ¼ 24.46; ommendations from dominant-brand combination users.
t(302) ¼ 4.81, p < .001, d ¼ 0.55); however, the differ- We confirmed this full account both by measuring the
ence was eliminated in the explicit conditions (MDominant ¼ mediators in study 2A and—for convergent evidence
60.31, SD ¼ 22.71 vs. MMixed ¼ 63.89, SD ¼ 21.24; t(296) (Spencer et al. 2005)—by manipulating the first media-
¼ 1.41, p ¼ .161, d ¼ 0.16; figure 9, right panel). tor in study 2B. Taken together, these results not only
FIGURE 9
DECISION THOUGHTFULNESS (LEFT) AND TRUST IN RECOMMENDATION (RIGHT) (STUDY 2B)
70 Mixed-brand 70 Mixed-brand
Dominant-brand Dominant-brand
Decision Thoughulness
60 60

50 50
40 40
30 30
20 20
Inferred Explicit Inferred Explicit
Note: Error bars indicate 95% confidence intervals.
FIGURE 10
MODERATED MEDIATION MODEL (STUDY 2B)
F(4, 597) = 121.73***

R2 = .45 Decision
a = 16.21*** Thoughtfulness b = .61***
(M2)
Dominant-
vs. Mixed-Brand
Combination (IV)
c’ = .75 Trustworthiness
X (DV)
Relative Importance of
brand: Inferred vs.
Moderated Mediation:
explicit (M1)
9.84, 95% CI [4.84, 15.08]
Note: Figure shows the unstandardized regression coefficients and the bootstrap 95% confidence interval of the indirect effect of the moderated mediation.
Dominant-brand combination was coded 1 and mixed-brand combination was coded 0. ***p < .001.
confirm our account of why the effect emerges but also precisely, less untrustworthy) by participants who were,
elucidate when the effect emerges (or does not emerge). themselves, endorsers of the dominant focal brand. To do
Specifically, the negative effect of dominant-brand com- this, study 3 tested a new product category, electronic devi-
binations can be “turned off” when observers explicitly ces, and used Apple as the focal brand. We measured both
know that brand did not play an outsized role in the participants’ trust in dominant- or mixed-brand combina-
choice. tion users and also their own attitudes toward Apple.
STUDY 3: BOUNDARY CONDITIONS Method

Study 3 examined circumstances in which dominant-
brand combination users may not be deemed less trustwor-
35.29 years, SD ¼ 10.41).
thy—even when observers are free to form their own infer-
ences. Prior research has shown that unfavorable Procedure. Study 3 used a two-condition, between-
perceptions of brand usage can be mitigated by observers’ subjects design. Participants were asked to imagine they en-
own attitude or affinity toward the brand (Ferraro et al. countered a person with five electronic items. To depict the
2013). Therefore, we tested whether dominant-brand com- array, participants saw a table with two columns: one that
bination users may be seen as more trustworthy (or, more listed the five items (laptop sleeves, smartwatch, headphones,
television box, tablets) and one that showed the brand logo of STUDY 4A: REACTION TO
each item. The dominant-brand combination user had all RECOMMENDATIONS IN BLOG POST
items from Apple, while the mixed-brand combination user
had items from all different brands. The focal item was the Study 4A tested the effect for recommendations about
Apple smartwatch. (Exact stimuli in web appendix.) baby gear items on a fictitious blog and recruited a highly
Using slider bars, we measured both general trust (“How involved sample: new parents.
much would you trust this consumer for recommendations
about electronics?”) and specific trust (“How much would Method

you trust this consumer for recommendations about
smartwatches?”). We also measured perceived typicality Participants. Participants were recruited from three
(“How likely is it that you could encounter a person like different subgroups (Pregnancy, Mom, and New Parent) on
this?”). Finally, we measured participants’ attitudes toward Reddit.com, a large online discussion forum. As an incen-
the focal brand (“How much do you like Apple tive, participants were eligible to enter a raffle for a $50
products?”). Amazon gift card. Data collection was set to stop on a pre-
specified date (four days after launch), with the goal of col-
lecting as many responses as possible during the window.
We collected data from 67 participants (91% female; Mage
Participants reported less trust in dominant-brand combi- ¼ 31.34 years, SD ¼ 5.05). Among those, 12% were cur-
nation users, both for general recommendations about elec- rently pregnant, and 91% already had children.
tronics (MDominant ¼ 54.59, SD ¼ 29.30 vs. MMixed ¼
Procedure. Participants were asked to imagine that
69.25, SD ¼ 21.76; t(200) ¼ 4.03, p < .001, d ¼ 0.57)
they wanted to buy a high chair, and that while they were
and for specific recommendations about the focal item
looking online, they found the blog of a mom who wrote a
(MDominant ¼ 55.99, SD ¼ 28.94 vs. MMixed ¼ 69.60, SD ¼
post about baby gear recommendations. To depict the rec-
23.93; t(200) ¼ 3.64, p < .001, d ¼ 0.51). There
ommendations, participants saw a table with two columns:
was no significant difference in perceived typicality
one that listed the four items (diaper bag, stroller organizer,
(MDominant ¼ 74.48, SD ¼ 22.88 vs. MMixed ¼ 71.37, SD ¼
play mat, high chair) and one that listed the brand of each
23.72; t(200) ¼ .95, p ¼ .344, d ¼ 0.13) or attitudes toward
item. The brands for the four baby gear items were either
the Apple brand (MDominant ¼ 56.82, SD ¼ 30.87 vs.
four different brands (Graco, Evenflo, Fisher Price,
MMixed ¼ 63.85, SD ¼ 30.38; t(200) ¼ 1.63, p ¼ .104,
Chicco; mixed-brand combination condition) or all Chicco
d ¼ 0.23).
(dominant-brand combination condition). The focal item
We next examined whether participants’ attitudes to-
was the Chicco high chair. (Exact stimuli in web
ward the focal brand moderated trust. For general trust, we
appendix.)
found a significant interaction between brand combination
We measured trust in recommendations for the focal
and brand attitude (t(198) ¼ 4.57, p < .001), suggesting
product: “How much would you trust this person for rec-
that—when participants’ attitude toward the brand is very
ommendations about high chairs?” (1 ¼ “Not at all” to 7 ¼
positive (72.24 or higher, based on the Johnson–Neyman
“Extremely”). Additionally, we measured participants’ fol-
test; Hayes 2018)—the difference in trust between domi-
lowing intention: “How likely or unlikely would you be to
nant- and mixed-brand combination users becomes nonsig-
follow this person on social media (e.g., Facebook,
nificant at p < .05; this was true for 42% of the sample.
Instagram, Snapchat)?” (1 ¼ “Very unlikely” to 7 ¼ “Very
For specific trust, the interaction was not significant
likely”). Finally, we observed participants’ actual click-
(t(198) ¼ 1.29, p ¼ .200); however, using the Johnson–
through behavior by examining the proportion of partici-
Neyman test, results show that at even higher levels of atti-
tude toward the brand (81.83 or higher), the difference in pants who chose to be redirected to the webpage to learn
trust between dominant- and mixed-brand combination more about the focal product: “Do you want to find out
users becomes nonsignificant at p < .05; in total, 32% of more about the Chicco high chair she recommends?”
the sample fell into that region (figure 11). Choices were labeled “YES, please send me to the web-
page showing the Chicco high chair she recommends” and
“NO, I am not interested in learning more about the Chicco
STUDIES 4A–4C: THE EFFECT ON high chair she recommends.” If a participant chose “yes,”
OBSERVERS’ OWN BEHAVIOR they were provided a link that took them to the webpage of
the Chicco high chair.
Having shown the main effect, explored the process, and
demonstrated an important boundary condition, studies
4A–4C examined the downstream consequences: how do
Results
observers’ inferences about dominant- versus mixed-brand Participants trusted the high chair recommendation less
combination users affect their own behavior? from a dominant-brand combination user (M ¼ 2.41, SD ¼
FIGURE 11
EFFECT OF BRAND COMBINATION ACROSS VARYING LEVELS OF OWN BRAND ATTITUDE (STUDY 3)
A General Trust B Specific Trust

90 90
Dominant-brand Dominant-brand

80 80
Mixed-brand Mixed-brand
70 70
60 60
50 50
40 40
30 30
M-1SD M M+1SD M-1SD M M+1SD
(29.58) (60.34) (91.09) (29.58) (60.34) (91.09)
Atude towards the brand Atude towards the brand
Note: Values in parentheses represent mean (þ/ 1SD) attitude toward the brand.
1.23) than a mixed-brand combination user (M ¼ 3.64, sample size in study 4B to better power our dichotomous
SD ¼ 1.45; t(65) ¼ 3.72, p < .001, d ¼ 0.91). behavioral dependent variable and ensure an adequate mix
Consequently, participants reported a lower intention to by gender.)
follow the dominant-brand combination user on social me-
dia (M ¼ 1.38, SD ¼ 0.95) than the mixed-brand combina-
tion user (M ¼ 2.48, SD ¼ 1.68; t(65) ¼ 3.32, p ¼ .001,
Procedure. Participants were asked to imagine going
d ¼ 0.81). In addition, they were less likely to opt to
on Instagram and seeing a post containing five clothing
learn more about the recommended product (Dominant ¼
items that an Instagram user recommends (example in fig-
8.8% vs. Mixed ¼ 24.2%, v2(1, N ¼ 67) ¼ 2.90, p ¼ .089,
ure 13; exact stimuli in web appendix). These five clothing
Cramer’s v ¼ .21, see figure 12); while only marginally
items were either all from different brands in the mixed-
significant, the effect size on the clickthrough behavior
brand combination condition (Zara, Levi’s, Mango,
was meaningful (i.e., small-to-medium Cramer’s v, c.f.
Benetton, Gap) or all from one brand in the dominant-
Kim 2017). Mediation analysis confirmed that the reduced
brand combination condition (Gap). The focal item was the
trust in recommendations for the dominant-brand combina-
Gap t-shirt. For female respondents, the post was made by
tion user significantly drove observers’ reduced intention
Amy and contained women’s clothing items; for male
to follow the user on social media (95% CI [0.95,
respondents, the post was made by Jack and contained
0.25]) and reduced clickthrough rates to learn more about
men’s clothing items. (Results were not moderated by gen-
the high chair (90% CI [2.70, 0.03], Hayes 2018).
der, and this factor was omitted from the below analysis.)
In both conditions, the Instagram post had the typical
STUDY 4B: REACTION TO “Learn More” banner at the bottom of the picture, which—
RECOMMENDATIONS IN SOCIAL MEDIA mimicking Instagram’s actual format—directs those who
POST click it to a webpage associated with the post’s content.
We used the same questions as in study 4A: (1) how
Study 4B replicated study 4A’s effect using Instagram
much they trust this person for their recommendation about
posts promoting a fashion blog.
the focal item; (2) how likely they are to follow this person
on social media, and (3) whether they wanted to be redir-
Method ected to a new webpage to learn more about the focal prod-
Participants. Participants were recruited from uct. As in study 4A, the latter question was a consequential
Amazon’s Mechanical Turk (N ¼ 500; 54% male; Mage ¼ behavioral measure—that is, saying “Yes” actually di-
35.43 years, SD ¼ 10.50). (While we aimed for 100 partici- rected participants to a link containing the Gap t-shirt’s
pants per cell in most previous studies, we increased webpage.
FIGURE 12
OBSERVERS’ REACTION TO RECOMMENDATIONS (STUDIES 4A AND 4B)

Note: Top shows results of study 4A (observers’ reaction to blog post); bottom shows results of study 4B (observers’ reaction to social media post). Error bars indi-
cate 95% confidence intervals.
Results STUDY 4C: TAKING A

The results conceptually replicated study 4A. RECOMMENDATION
Participants trusted the t-shirt recommendations less from
dominant-brand combination users (MDominant ¼ 35.13, Study 4C used a blog entry and asked participants to
SD ¼ 26.28 vs. MMixed ¼ 41.62, SD ¼ 26.97; t(498) ¼ make a consequential product choice with an option to
2.72, p ¼ .007, d ¼ 0.24). Consequently, participants take—or forgo—someone else’s recommendation.
reported a lower intention to follow the dominant-brand
combination user on social media (MDominant ¼ 21.92, Method
SD ¼ 27.02 vs. MMixed ¼ 28.84, SD ¼ 30.46; t(498) ¼
2.69, p ¼ .007, d ¼ 0.24), and were significantly less Participants. Participants were recruited from
likely to opt to learn more about the recommended product Amazon’s Mechanical Turk (N ¼ 399; 45% male; Mage ¼
(Dominant ¼ 10.8% vs. Mixed ¼ 17.3%, v2(1, N ¼ 500) ¼ 36.22 years, SD ¼ 11.42). (As was true for study 4B, we
4.40, p ¼ .036, Cramer’s v ¼ .09; see figure 12). increased the sample size in study 4C to provide better
Mediation analysis confirmed that the reduced trust for the power for our dichotomous dependent variable.)
dominant-brand combination user (vs. mixed-brand combi-
nation user) drove observers’ reduced intention to follow Procedure. Participants began by reporting demo-
the user on social media (95% CI [9.14, 1.32]) and re- graphic information (gender, age, and country of resi-
duced clickthrough rates to learn more about the Gap t- dence), which allowed us to match the stimuli to their self-
shirt (95% CI [0.76, 0.12], Hayes 2018). reported gender.
FIGURE 13
EXAMPLE STIMULI SHOWN TO FEMALE PARTICIPANTS IN STUDY 4B

Note: Dominant-brand combination on left, mixed-brand on right. Focal item: gap t-shirt.
Next, all participants read a blog post in which the blog- recommendations about running shoes?” (1 ¼ “Not at all”
ger had curated six running products for the winter season to 7 ¼ “Extremely”). Furthermore, to examine the effect
(see figure 14 for sample stimuli); self-identified males on recommendation taking, participants were told that they
saw men’s running gear items, while self-identified had the chance to enter one of two raffles: (1) Raffle A for
females saw women’s running gear items. Beneath the “an item that this blog writer recommends worth $40” or
blog text, participants saw pictures of the six running prod- (2) Raffle B for “an item of your choice worth $20.”
ucts (shoes, headband, shirt, pants, socks, and jacket) in a Participants indicated which raffle they preferred to enter.
style designed to reflect a typical manner in which recom- (One person was randomly selected to receive either a $20
mended products may be displayed on blogs. Across both or $40 prize, depending on their choice.) We measured the
conditions, all participants saw the exact same product pic- proportion of participants who chose the more valuable
tures; however, in one condition (dominant-brand combi- “Raffle A” as a proxy for recommendation-taking.
nation), all items were from Adidas (i.e., had the Adidas Finally, participants indicated how many times per
brand logo on them), whereas in the other condition month they run (times/month: M ¼ 4.93, SD ¼ 6.83,
(mixed-brand combination), only the shoe was from Median ¼ 2.00).
Adidas and different brand logos appeared on the five non-
focal items (Odlo headband, Under Armour shirt, Puma Results
pants, Nike socks, and Salomon jacket). Thus, the Adidas Replicating all previous studies, participants were less
shoe was the focal item that was identical across condi- likely to trust the dominant-brand combination blogger
tions. For female and male participants, the products varied (M ¼ 4.25, SD ¼ 1.54) than the mixed-brand combination
in their shape/color but were closely matched on style; the blogger (M ¼ 4.70, SD ¼ 1.19; t(397) ¼ 3.26, p ¼ .001, d
results were not moderated by gender, and this factor has ¼ 0.33). In addition, fewer participants chose to enter the
been omitted from the below analysis. (Exact stimuli in objectively more valuable Raffle A—for an item the blogger
web appendix). recommended—when the recommendation would come from
Next, we measured trust in recommendations for the fo- a dominant-brand combination blogger (43.3%) versus a
cal product: “How much would you trust this person for mixed-brand combination blogger (55.1%; v2(1, N ¼ 399) ¼
FIGURE 14
EXAMPLE STIMULI SHOWN TO MALE PARTICIPANTS IN STUDY 4C

Note: Dominant-brand combination on left, mixed-brand on right. Focal item: Adidas shoes.
5.53, p ¼ .019, Cramer’s v ¼ .12). Examined differently, par- dominant-brand combination user in a consequential
ticipants in the dominant-brand combination condition were choice task (study 4C).
marginally less likely than chance (vs. 50%; p ¼ .067) to take
the raffle with the blogger’s recommendation—perhaps sug- GENERAL DISCUSSION
gestive of active avoidance rather than passive indifference.
Mediation analysis confirmed that the reduced trust in recom- Nine studies examined how different brand combina-
mendations for the dominant-brand combination user (vs. tions—dominant or mixed—affect observers’ inferences
mixed-brand combination user) significantly drove observers’ about a target consumer’s purchase decisions, and in turn,
reduced likelihood to take the raffle with the dominant-brand trust in the target consumer’s recommendations.
combination blogger’s recommendation (95% CI [0.36, Specifically, observers tended to trust recommendations
0.07], Hayes 2018). from dominant-brand combination users less than those
from mixed-brand combination users (studies 1A–1C).
This was driven by the broader inferences made about the
Studies 4A–4C Discussion underlying purchase decision: observers tended to believe
Taken together, studies 4A–4C replicated the main ef- that dominant-brand combination users had placed rela-
fect and demonstrated how this reduced trust impacts tively greater importance on the brand—a feature that often
behaviors and intentions for the observers themselves. serves as a heuristic for consumers (Hoyer and Brown
Observers were less likely to choose to learn more about 1990; Maheswaran et al. 1992; Simonson and Winer
the same focal product when it was recommended by a 1992)—and therefore believed dominant-brand combina-
dominant-brand combination user (studies 4A and 4B), tion users had been less thoughtful in deciding which item
were less likely to want to follow a dominant-brand combi- to buy (studies 2A–2B). The effect was mitigated when
nation user on social media (studies 4A and 4B), and were observers had particularly positive attitudes toward the fo-
less likely to actively accept the recommendation of a cal brand (study 3). Overall, we found evidence of this
effect across an array of product categories (running gear, that signals differ by combination—important differences
clothing, toiletries, electronic devices, baby gear) and stim- exist between our paradigms, mechanisms, and, accord-
uli (Instagram posts, product images, logos, product lists, ingly, the broader takeaways for practice.
blog posts), and also found evidence that these inferences First and foremost, our paradigm is specific to brands,
can affect observers’ own decision-making behavior (stud- not flavors—two features that researchers have empirically
ies 4A–4C). documented are distinct (Harlam and Lodish 1995; Inman
2001; Simonson and Winer 1992). For example, while
Theoretical Contribution shopping for yogurts, consumers tend to seek a variety of

flavors but only a single brand (Harlam and Lodish 1995;
From a theoretical perspective, we contribute to the liter-
Simonson and Winer 1992), which can be explained by dif-
ature on peer recommendations and word of mouth
ferences in satiation (i.e., people seek more variety for
(Bearden and Etzel 1982; Berger 2014; Bikhchandani et al.
sensory-related products like flavors; Inman 2001) and the
1992; Childers and Rao 1992; Ward 1974; Watts and
relative stability of preferences (i.e., brand preferences
Dodds 2007). More specifically, we identify an important
change less frequently than flavor preferences; Inman
cue that can increase (or decrease) recommender trust—ir-
2001; Simonson and Winer 1992). Furthermore, brands
respective of who the recommender actually is—and thus
have properties that more one-dimensional attributes lack:
add to a nascent stream of research investigating the impact
consumers have relationships with brands (Fournier and
of others’ (actual or inferred) decision processes on one’s
Yao 1997), derive meaning and identity from brands
own perceptions and behavior. Previous research has
(Aaker 1997; Belk 1988; Escalas and Bettman 2005;
shown that when people directly observe someone else’s
Keller 1993), and deploy brands as status signals (Ferraro,
decision process—and can see that an appropriate amount
Kirmani, and Matherly 2010; Ferraro et al. 2013). The
of effort and time was devoted to the choice—they are
same cannot be said for attributes such as flavor; consum-
more open to recommendations from that person (Kupor et
ers are seldom described as “flavor loyalists,” and the
al. 2014). Consistent with these findings, we demonstrate
“conspicuous usage of flavors” is unlikely to draw the
that—even when not privy to explicit information about
same ire as the conspicuous usage of brands (Ferraro et al.
the decision process—observers use brand combinations to
2013).
infer decision thoughtfulness, which affects trust in recom-
Imbued with such meaning, brands therefore act as a
mendations and observers’ own behavior.
“knowledge structure that can operate as a judgmental heu-
As importantly, we contribute to the branding literature
ristic” (Maheswaran et al. 1992, 318). This notion was em-
by investigating brand combinations, a “largely ignored
pirically corroborated by Simonson and Winer (1992), who
area of branding research” (Rahinel and Redden 2013,
went on to suggest that “the simplest choice heuristic. . . is
1291). While many studies have examined how the use of
to focus on the shelf section with the consumer’s favorite
a single brand affects observers’ inferences (Ferraro et al.
brand” (136)—to revert, in other words, to a dominant-
2013; Han et al. 2010; Rice et al. 2012), virtually none
brand assortment. Our entire account is predicated on pre-
have examined the effect of multiple combinations of
cisely this idea—of brands as heuristics—and indeed, this
brands. However, as our “Motivating the Inquiry” section
lies at the heart of our contribution with distinct implica-
revealed, people use, encounter, and notice different com-
tions for practice. Sela et al. (2019) do not predict such a
binations of brands across product categories. Furthermore,
relationship, nor would we predict that our model holds for
our studies empirically demonstrate that these combina-
flavors. Specifically, we would not expect—nor do we em-
tions, in turn, can materially change observers’ infer-
pirically find (supplementary study S8)—that flavors act as
ences—over and above the effects of single brands—
comparable heuristics, such that “dominant-flavor” deci-
suggesting that brand combinations merit further investiga-
sions are perceived as having been less thoughtfully made.
tion as a distinct unit of analysis.
Thus, our mechanism is not only distinct but it also pro-
The idea that combinations can affect inferences has
vides a novel and theoretically important contribution: that
been shown in a few other places (Rahinel and Redden
the extent to which a decision is seen as “brand-driven”
2013; Sela et al. 2019); our work complements and extends
may also affect inferences about how thoughtfully it was
this existing work. Of particular relevance, Sela et al.
made.
(2019) investigated how a product assortment’s variety can
influence observers’ judgments of expertise. By manipulat-
ing the variety of flavors within a single product type (e.g.,
Managerial Implications
many or few flavors of chocolates, coffee, or beer), they Beyond the academic contributions, these results also
found that variety-seeking can both increase and decrease have meaningful practical and managerial implications.
the perceived expertise of a target consumer as a function Our consumer surveys suggest that most people already no-
of the observer’s own expertise. While our work offers tice brand combinations; however, such combinations are
complementary insights—insofar as we also find evidence only likely to become more observable thanks to social
media and other technologies. For example, Amazon and to mitigate the risk of being overshadowed? Moreover, it is
others have introduced artificial intelligence-enabled apps unclear what the net effect of dominant- versus mixed-
that allow consumers to discover the brands other people brand combinations is. For instance, does the benefit of
are wearing simply by uploading fashion photos and then brand visibility from a dominant-brand combination out-
linking to products (Hanbury 2019), making brand combi- weigh the drawbacks of reduced trust in a recommender?
nations observable even where they would not be These are unanswered questions, and we are certainly not
otherwise. advocating for firms to adopt a blanket policy of actively
All of this is likely to be particularly relevant within the positioning themselves alongside competitors. However,

increasingly important context of influencer marketing our investigation suggests that the fear and active avoid-
(Burns 2016; Freberg et al. 2011; Gong et al. 2017). The ance of mixed-brand combinations may be exaggerated, at
most successful, trusted influencers are those who appear least as it pertains to observers’ trust. As our empirical
genuine (MacDonald 2016); thus, knowing how observing package suggests, there may be a signal benefit to embed-
consumers will perceive others’ product decisions is im- ding one brand within the context of others, a finding gen-
portant. Sometimes influencers explicitly reveal their brand erally consistent with “brand flirting” research highlighting
sponsorship, in which case they may be perceived more some benefits of consuming multiple brands (Aggarwal
like a paid endorser than a peer (which, as discussed above,
and Shi 2018; Consiglio et al. 2018). Therefore, we simply
falls outside the purview of this investigation); however,
suggest that—particularly as applied to influencer market-
influencers often skirt such disclosures and appear as ev-
ing—consuming multiple brands can make a consumer ap-
eryday people who have made their own (unsponsored)
pear as though they have made a thoughtful, trustworthy
product choices (Sah et al. 2018). Among this latter group,
our investigation highlights one such factor affecting influ- choice, which may actually improve observers’ responses.
encers’ trustworthiness—the brand combinations they have
supposedly chosen to consume—and speaks to the broader Limitations and Future Research
question: Holding the actual influencer personality con- We hope future research will investigate many other im-
stant, how should brands best position themselves—quite portant aspects of brand combinations and propose four
literally—in order to maximize the influencer’s influence? factors that may prove fruitful for further inquiry: brand
Perhaps unsurprisingly, managers prefer that influencers status, perceived brand dominance, brand-product fit, and
stick to dominant-brand combinations—for example, do combination cohesiveness.
not post photos of their brands alongside other brands. For First, our studies intentionally used combinations of
instance, one survey revealed that the second most com- brands that were of relatively comparable status and price
mon reason a firm-influencer partnership agreement cannot points in order to keep the broader combination relatively
be reached—right behind failure to finalize compensation neutral. However, one could see how—even within mixed-
terms—is involvement with a competing brand (Activate brand combinations—varying the status of brands might
2019). Relatedly, another marketing firm recently launched also elicit different reactions; for example, a combination
an artificial intelligence-based “conflict assessment tool”
containing Nike shoes, Asics shorts, Puma shirt, and
to ensure that influencers have not worked with or en-
Brooks hydration pack might be more well-regarded than a
dorsed a competing brand (B&T Magazine 2018). We con-
combination containing Nike shoes and an assortment of
firmed this preference ourselves by partnering with an
lower-status private label brands. Future research could fo-
influence marketing agency to survey 34 marketing man-
agers (Mage ¼ 38.26, SD ¼ 9.98; 82% women; median cus on the effects of brand status within combinations.
work experience ¼ 11 years; preregistered at https://aspre- Second, study 1B demonstrated that a combination need
dicted.org/tn9m3.pdf; full details in supplementary study not be exclusively comprised of a single brand to be less
S7). Managers were more satisfied with dominant- (vs. trusted; observers also reported lower trust in others who
mixed-) brand combination influencers (p < .001, used combinations in which a brand was repeated five or
d ¼ 1.31), believed their recommendations would be more six times out of seven. Therefore, when a large majority of
trustworthy (p ¼ .087, d ¼ 0.60), and were more likely to items in a combination comes from a single brand, it is
want to continue working with them in the future (p ¼ likely to appear dominant—thus inciting the effect we have
.001, d ¼ 1.23). As one manager explained, “I think [a documented. This leads to other questions about how a
mixed-brand combination] makes the experience more re- brand may come to be perceived as dominant within a
alistic. I just do not want people to recognize other combination. For instance, future work could further cali-
brands.” brate dominance on the basis of repetition (e.g., the number
This sentiment underscores the downside of mixed- of times a brand repeats in combinations of different sizes)
brand combinations: that competing brands get “air time” or examine dominance on the basis of other focalizing
alongside a firm’s own. Which brand(s) will observing aspects (e.g., the size of the brand logo, the status of the
consumers attend to most? Is it possible for a focal brand brand, etc.).
Third, our studies used stimuli in which the brand- our theory rests upon one-time, “snapshot” encounters, in
product fit was either strong (e.g., Apple and smart- which observers see a target consumer—and his or her
watches; Nike and running shoes), neutral (e.g., Gap and combination of brands—at a single point in time. This
work bags), or in some cases, unclear (e.g., “Brand A” and invites questions about how such inferences may change
work bags)—but never explicitly weak. However, it is pos- over time—for instance, how does loyalty to a single brand
sible that inferences might change if the combination in- over time compare to dominant-brand combinations?
cluded brands with actively weak or unexpected product fit Stepping back, people frequently use combinations of
(e.g., Ferrari-branded work bag). Additional work may ex- brands in everyday life. This research demonstrated why it

amine how brand-product fit within a combination affects is important to consider this broader unit of analysis: brand
perceptions. combinations can meaningfully change observer infer-
Fourth, we studied combinations in which products were ences, over and above single brands. While much opportu-
all from the same general category (e.g., running gear, toi- nity remains to gain a nuanced understanding of the
letries). However, future research could examine how dif- signaling effect of brand combinations, we hope the pre-
ferent levels of combination cohesiveness and coordination sent work will encourage future research in this direction.
affect inferences. For instance, one could imagine that all
Nike running products were created to “work together” DATA COLLECTION INFORMATION
(Rahinel and Redden 2013), though it is less likely that
same link holds between Nike running shoes, golf clubs, Both authors collected and analyzed the data of all stud-
and snowboards. Likewise, combinations could be more ies. Studies 1–3, 4B–4C, the two pretests with studies 1A
broadly construed and need not necessarily come from the and 2A, and the supplementary studies S2–S6 and S8 (web
same category at all; for instance, an observer might simul- appendix) were collected on Amazon Mechanical Turk.
taneously notice another consumer’s brand of shoes, watch, For study 4A, participants were invited via a post on three
and car—which may be encoded as its own brand combi- thematic subgroups—pregnant, moms, new parents—on
nation. Therefore, it is possible that dominant-brand com- Reddit (a large US online discussion forum). Participants
bination users will be judged differently depending on the for supplementary study S3 (Consumer Sample) and S7
(perceived or actual) coordination and cohesion of the were collected in collaborating with trnd Espa~na, a large
products within the combination. European Influence marketing agency. The AI-brand rec-
Further investigating these four factors (among many ognition data were collected in collaboration with bandi-
potential others) will deepen our understanding of brand do.run. The Instagram post data were collected by two
combinations, as well as help establish a fuller view of the research assistants. Studies 1C and 4A, as well as the sup-
boundaries of the effect—boundaries that undoubtedly ex- plementary studies S4 and S5, were collected in the aca-
ist. (For instance, perhaps high-status, dominant-brand demic year 2017–2018; studies 1A, 1B, 2A, 3, and the
combinations would be more favorably perceived than supplementary study S6 in the academic year 2018–2019;
low-status, mixed-brand combinations.) Additionally, the Pretest with study 2A, studies 2B, 4B, and the supple-
follow-up work should also consider other limitations of mentary studies S3, S7, and S8 in the academic year 2019–
the present work. For example, our studies made brands in- 2020; and the AI-brand recognition study (supplementary
tentionally salient to varying degrees; however, this is not studies S1), the Instagram post study (supplementary stud-
always the case in more naturalistic settings. Even when ies S2), the brand usage survey, the Pretest with study 1A,
logos or brand names appear across a given set of products, study 4C, and the Typicality Pretests in Spring/Fall 2020.
it is not guaranteed that observers will notice all—or even The data is currently stored on a shared Dropbox folder
any—of them. Perhaps, for instance, consumers are more and on OSF (https://osf.io/vyfz3).
likely to notice—or even actively seek out—brand combi-
nations when they are “in the market” for a particular prod-
uct category. Thus, when considering the external REFERENCES
application of this work, it will be important to establish Aaker, Jennifer L. (1997), “Dimensions of Brand Personality,”
when (and why) consumers are more or less likely to notice Journal of Marketing Research, 34 (3), 347–56.
brand combinations (vs. single brands or no brands at all). Activate (2019), “Double or Nothing: Betting Big on Influencer
Relatedly, our findings do not allow us to deduce the im- Marketing,” Activate, May 9, https://try.activate.social/dou-
portance of brand combinations relative to other types of ble-or-nothing-betting-big-on-influencer-marketing.
cues. Future research might investigate interesting follow- Aggarwal, Pankaj and Mengze Shi (2018), “Monogamous Versus
up questions, such as: In a richer environment, with many Polygamous Brand Relationships,” Journal of the
Association for Consumer Research, 3 (2), 188–201.
competing cues, how much relative importance is given to B&T Magazine (2018), “Hypetap Unveils New AI-Driven Brand
brand combinations? And how might brand combinations Safety Products,” http://www.bandt.com.au/marketing/hype-
interact with other types of cues—such as those pertaining tap-unveils-new-ai-driven-brand-safety-products. Accessed
to the personal traits of the recommender? Furthermore, September 10, 2021.
Barasz, Kate, Tami Kim, and Ioannis Evangelidis (2019), “I Know Consumer-Brand Relationships,” International Journal of
Why You Voted for Trump: (over)Inferring Motives Based Research in Marketing, 14 (5), 451–72.
on Choice,” Cognition, 188, 85–97. Freberg, Karen, Kristin Graham, Karen McGaughey, and Laura A.
Bearden, William O. and Michael J. Etzel (1982), “Reference Freberg (2011), “Who Are the Social Media Influencers? A
Group Influence on Product and Brand Purchase Decisions,” Study of Public Perceptions of Personality,” Public Relations
Journal of Consumer Research, 9 (2), 183–94. Review, 37 (1), 90–2.
Belk, Russell W. (1988), “Possessions and the Extended Self,” Giffin, Kim (1967), “The Contribution of Studies of Source
Journal of Consumer Research, 15 (2), 139–68. Credibility to a Theory of Interpersonal Trust in the
Berger, Jonah (2014), “Word of Mouth and Interpersonal Communication Process,” Psychological Bulletin, 68 (2),

Communication: A Review and Directions for Future 104–20.
Research,” Journal of Consumer Psychology, 24 (4), Gong, Shiyang, Juanjuan Zhang, Ping Zhao, and Xuping Jiang
586–607. (2017), “Tweeting as a Marketing Tool: A Field Experiment
Bikhchandani, Sushil, David Hirshleifer, and Ivo Welch (1992), in the TV Industry,” Journal of Marketing Research, 54 (6),
“A Theory of Fads, Fashion, Custom, and Cultural Change as 833–50.
Informational Cascades,” Journal of Political Economy, 100 Hamilton, Ryan, Kathleen D. Vohs, and Ann L. McGill (2014),
(5), 992–1026. “We’ll Be Honest, This Won’t Be the Best Article You’ll
Burns, Kelli S. (2016), “How the Top Social Media Brands Use Ever Read: The Use of Dispreferred Markers in Word-of-
Influencer and Brand Advocacy Campaigns to Engage Fans,” Mouth Communication,” Journal of Consumer Research, 41
in Public Relations and Participatory Culture Fandom, (1), 197–212.
Social Media and Community Engagement, ed. Amber Han, Young Jee, Joseph C. Nunes, and Xavier Drèze (2010),
Hutchins and Natalie T. J. Tindallessay, New York, NY: “Signaling Status with Luxury Goods: The Role of Brand
Taylor and Francis, 58–70. Prominence,” Journal of Marketing, 74 (4), 15–30.
Busch, Paul and David T. Wilson (1976), “An Experimental Hanbury, Mary (2019), “Amazon Is Launching the Shazam for
Analysis of a Salesman’s Expert and Referent Bases of Fashion, Which Finds Clothes You Want Simply by
Social Power in the Buyer-Seller Dyad,” Journal of Analyzing a Photo,” Business Insider, June 6, https://www.
Marketing Research, 13 (1), 3–11. businessinsider.com/amazon-launches-shazam-for-clothes-
Carl, Walter J. (2008), “The Role of Disclosure in Organized 2019-6. Accessed September 10, 2021.
Word-of-Mouth Marketing Programs,” Journal of Marketing Harlam, Bari A. and Leonard M. Lodish (1995), “Modeling
Communications, 14 (3), 225–41. Consumers’ Choices of Multiple Items,” Journal of
Childers, Terry L. and Akshay R. Rao (1992), “The Influence of Marketing Research, 32 (4), 404–18.
Familial and Peer-Based Reference Groups on Consumer Hayes, Andrew F. (2018), Introduction to Mediation, Moderation,
Decisions,” Journal of Consumer Research, 19 (2), 198–211. and Conditional Process Analysis: A Regression-Based
Cialdini, Robert B. (2008), Influence: Science and Practice, Approach, New York, NY: The Guilford Press.
Boston, MA: Person Education. Hovland, Carl I., Irving L. Janis, and Harold H. Kelley (1953),
Consiglio, Irene, Daniella M. Kupor, Francesca Gino, and Michael Communication and Persuasion; Psychological Studies of
I. Norton (2018), “Brand (in)Fidelity: When Flirting with the Opinion Change, New Haven, CT: Yale University Press.
Competition Strengthens Brand Relationships,” Journal of Hoyer, Wayne D. and Steven P. Brown (1990), “Effects of Brand
Consumer Psychology, 28 (1), 5–22. Awareness on Choice for a Common, Repeat-Purchase
Doney, Patricia M. and Joseph P. Cannon (1997), “An Product,” Journal of Consumer Research, 17 (2), 141–8.
Examination of the Nature of Trust in Buyer–Seller Hsiao, Kuo-Lun, Judy Chuan-Chuan Lin, Xiang-Ying Wang, Lu
Relationships,” Journal of Marketing, 61 (2), 35–51. Hsi-Peng, and Yu Huei-Ju (2010), “Antecedents and
Edwards, Jeffrey R. and Lisa S. Lambert (2007), “Methods for Consequences of Trust in Online Product Recommendations:
Integrating Moderation and Mediation: A General Analytical An Empirical Study in Social Shopping,” Online Information
Framework Using Moderated Path Analysis,” Psychological Review, 34 (6), 935–53.
Methods, 12 (1), 1–22. Inman, J. Jeffrey (2001), “The Role of Sensory-Specific Satiety in
Escalas, Jennifer E. and James R. Bettman (2005), “Self- Attribute-Level Variety Seeking,” Journal of Consumer
Construal, Reference Groups, and Brand Meaning,” Journal Research, 28 (1), 105–20.
of Consumer Research, 32 (3), 378–89. Keller, Kevin L. (1993), “Conceptualizing, Measuring, and
Evans, Anthony M. and Philippe P. F. M. van de Calseyde (2017), Managing Customer-Based Brand Equity,” Journal of
“The Effects of Observed Decision Time on Expectations of Marketing, 57 (1), 1–22.
Extremity and Cooperation,” Journal of Experimental Social Kim, Hae-Young (2017), “Statistical Notes for Clinical
Psychology, 68, 50–9. Researchers: Chi-Squared Test and Fisher’s Exact Test,”
Ferraro, Rosellina, Amna Kirmani, and Ted Matherly (2010), Restorative Dentistry & Endodontics, 42 (2), 152–5.
“Signaling Identity through Brands: The Role of Perceived Kupor, Daniella and Zakary Tormala (2018), “When Moderation
Authenticity,” in NA—Advances in Consumer Research Fosters Persuasion: The Persuasive Power of Deviatory
Volume 37, ed. Margaret C. Campbell, Jeff Inman, and Rik Reviews,” Journal of Consumer Research, 45 (3), 490–510.
Pieters, Duluth, MN: Association for Consumer Research, Kupor, Daniella M., Zakary L. Tormala, Michael I. Norton, and
81–4. Derek D. Rucker (2014), “Thought Calibration: How
——— (2013), “Look at Me! Look at Me! Conspicuous Brand Thinking Just the Right Amount Increases One’s Influence
Usage, Self-Brand Connection, and Dilution,” Journal of and Appeal,” Social Psychological and Personality Science,
Marketing Research, 50 (4), 477–88. 5 (3), 263–70.
Fournier, Susan and Julie L. Yao (1997), “Reviving Brand Lee, Kyung-Tag and Dong-Mo Koo (2012), “Effects of Attribute
Loyalty: A Reconceptualization within the Framework of and Valence of e-WOM on Message Adoption: Moderating
Roles of Subjective Knowledge and Regulatory Focus,” Rice, Dan H., Katie Kelting, and Richard J. Lutz (2012), “Multiple
Computers in Human Behavior, 28 (5), 1974–84. Endorsers and Multiple Endorsements: The Influence of
MacDonald, Kelley (2016), “Study: Genuine Messaging Is the #1 Message Repetition, Source Congruence and Involvement on
Driver of Engagement on Branded Content,” Bloglovin, Brand Attitudes,” Journal of Consumer Psychology, 22 (2),
https://influence.bloglovin.com/study-genuine-messaging-is- 249–59.
the-1-driver-of-engagement-on-branded-content-ce262e51f5ce. Sah, Sunita, Prashant Malaviya, and Debora Thompson (2018),
Accessed September 10, 2021. “Conflict of Interest Disclosure as an Expertise Cue:
Maheswaran, Durairaj, Diane M. Mackie, and Shelly Chaiken Differential Effects Due to Automatic Versus Deliberative
(1992), “Brand Name as a Heuristic Cue: The Effects of Task Processing,” Organizational Behavior and Human Decision

Importance and Expectancy Confirmation on Consumer Processes, 147, 127–46.
Judgments,” Journal of Consumer Psychology, 1 (4), 317–36. Schrift, Rom Y. and Moty Amar (2015), “Pain & Preferences:
Mills, Judson (1966), “Opinion Change as a Function of the Observed Decisional Conflict & the Convergence of
Communicator’s Desire to Influence and Liking for the Preferences,” Journal of Consumer Research, 42 (2),
Audience,” Journal of Experimental Social Psychology, 2 515–34.
Sela, Aner, Liat Hadar, Siân Morgan, and Michal Maimaran
(2), 152–9.
(2019), “Variety-Seeking and Perceived Expertise,” Journal
Mills, Judson and Elliot Aronson (1965), “Opinion Change as a
of Consumer Psychology, 29 (4), 671–9.
Function of the Communicator’s Physical Attractiveness and
Simonson, Itamar and Russell S. Winer (1992), “The Influence of
Desire to Influence,” Journal of Personality and Social
Purchase Quantity and Display Format on Consumer
Psychology, 1 (2), 173–7. Preference for Variety,” Journal of Consumer Research, 19
Mills, Judson and John Harvey (1972), “Opinion Change as a (1), 133–8.
Function of When Information about the Communicator is Spencer, Steven J., Mark P. Zanna, and Geoffrey T. Fong (2005),
Received and Whether He Is Attractive or Expert,” Journal “Establishing a Causal Chain: Why Experiments Are Often
of Personality and Social Psychology, 21 (1), 52–5. More Effective than Mediational Analyses in Examining
Moorman, Christine, Rohit Deshpande, and Gerald Zaltman Psychological Processes,” Journal of Personality and Social
(1993), “Factors Affecting Trust in Market Research Psychology, 89 (6), 845–51.
Relationships,” Journal of Marketing, 57 (1), 81–101. Ward, Scott (1974), “Consumer Socialization,” Journal of
Pornpitakpan, Chanthika (2004), “The Persuasiveness of Source Consumer Research, 1 (2), 1–14.
Credibility: A Critical Review of Five Decades’ Evidence,” Watts, Duncan J. and Peter Sheridan Dodds (2007), “Influentials,
Journal of Applied Social Psychology, 34 (2), 243–81. Networks, and Public Opinion Formation,” Journal of
Rahinel, Ryan and Joseph P. Redden (2013), “Brands as Product Consumer Research, 34 (4), 441–58.
Coordinators: Matching Brands Make Joint Consumption Wernerfelt, Birger (1990), “Advertising Content When Brand
Experiences More Enjoyable,” Journal of Consumer Choice Is a Signal,” The Journal of Business, 63 (1),
Research, 39 (6), 1290–9. 91–8.

2c. Engeler, I., Barasz, K. (2021) - From Mix-and-Match To Head-to-Toe

Uploaded by

Copyright:

Available Formats

2c. Engeler, I., Barasz, K. (2021) - From Mix-and-Match To Head-to-Toe

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

2c. Engeler, I., Barasz, K. (2021) - From Mix-and-Match To Head-to-Toe

Uploaded by

Copyright:

Available Formats

From Mix-and-Match to Head-to-Toe: How

Brand Combinations Affect Observer Trust

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Consumers use brands in many combinations, from mixing-and-matching multiple

Keywords: choice inferences, brand combinations, decision thoughtfulness, con-

C onsumers use brands in many combinations—from

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

TRIANGULATION OF PREVALENCE OF REPEATED BRAND USAGE—EXAMPLE OF ACTIVEWEAR.

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

SURVEY STIMULI AND RESULTS

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

observers must believe a purchase decision was made; Method

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

FIGURE 3 seen as atypical and thus untrustworthy—by directly mea-

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

TABLE 2 combination is mostly comprised of one repeated brand,

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

FIGURE 5 brand combination users. This occurred using Instagram

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

BRAND COMBINATIONS DISPLAYED IN PRETEST AND STUDY 2A

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

FIGURE 7 (dominant-brand combination) or all from different brands

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

MEDIATION ANALYSIS (STUDY 2A)

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Total indirect effect:

Partitioning of total indirect effect:

DECISION THOUGHTFULNESS (LEFT) AND TRUST IN RECOMMENDATION (RIGHT) (STUDY 2B)

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Note: Error bars indicate 95% confidence intervals.

MODERATED MEDIATION MODEL (STUDY 2B)

F(4, 597) = 121.73***

STUDY 3: BOUNDARY CONDITIONS Method

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

A General Trust B Speciﬁc Trust

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

OBSERVERS’ REACTION TO RECOMMENDATIONS (STUDIES 4A AND 4B)

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Results STUDY 4C: TAKING A

EXAMPLE STIMULI SHOWN TO FEMALE PARTICIPANTS IN STUDY 4B

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

EXAMPLE STIMULI SHOWN TO MALE PARTICIPANTS IN STUDY 4C

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

Downloaded from https://academic.oup.com/jcr/article/48/4/562/6353610 by Université de Lausanne user on 07 October 2022

You might also like