You are on page 1of 14

Article

Journal of Marketing
2020, Vol. 84(1) 52-65
Featuring Mistakes: The Persuasive Impact ª American Marketing Association 2019
Article reuse guidelines:

of Purchase Mistakes in Online Reviews sagepub.com/journals-permissions


DOI: 10.1177/0022242919882428
journals.sagepub.com/home/jmx

Taly Reich and Sam J. Maglio

Abstract
Companies often feature positive consumer reviews on their websites and in their promotional materials in an attempt to
increase sales. However, little is known about which particular positive reviews companies should leverage to optimize sales.
Across four lab studies involving both hypothetical and real choices as well as field data from a retailer’s website (Sephora), the
authors find that consumers are more likely to purchase a product if it is recommended by a reviewer who has (vs. has not) made
a prior purchase mistake. The authors define a purchase mistake as a self-identified suboptimal decision whereby people purchase
a product that subsequently fails to meet a threshold level of expected performance. This persuasive advantage emerges because
consumers perceive reviewers who admit a purchase mistake as having more expertise than even reviewers whose purchase
experience has not been marred by mistakes. As a result, in marketers’ attempts to increase the persuasive influence of reviews
featured in their promotional materials, they may inadvertently decrease it by omitting the very information that would lead
consumers to be more likely to purchase recommended products.

Keywords
online reviews, persuasion, mistakes, expertise, word of mouth
Online supplement: https://doi.org/10.1177/0022242919882428

Potential purchasers place a great deal of stock in product but rather how the review conveyed its positivity: in explicitly
reviews written by previous purchasers. More and more, these stating that they endorse the product, online reviews are more
product reviews are written, read, and evaluated online (Lips- likely to lead to purchases than those with more implicit posi-
man et al. 2019). Online reviews gain as much trust as personal tivity. This more nuanced examination of the content of online
recommendations for the majority of consumers (85%; Murphy reviews—going beyond the mere valence of the review—opens
2017), and a glowing review motivates behavior more than the door to complementary approaches that ask not whether
discounts and other offers (in the domains of durable goods reviewers like the product but what the reviewers say when
and electronics; Kats 2018). Consumers seem to want product expressing their positivity or negativity, providing insight as
reviews (www.iperceptions.com), and firms seem happy to to how firms can optimize the persuasive influence of featuring
offer such reviews on their websites, in their advertising, and reviews. To illustrate how we propose to add to this growing
elsewhere (Arcillo 2011; Dellarocas 2003; Harper 2014; Raw- and important literature, consider the following two real
lins 2011). Indeed, firms are increasingly engaging in efforts reviews from Amazon:
around “review solicitation” and “online reputation manage-
ment,” incentivizing previous purchasers to write reviews in When my first Canon battery expired, I purchased a knock-off.
exchange for discounted or free products. What a mistake. It lasted about 1/3 as long as the Canon. (Richard
As firms invest in, and consumers trust, online reviews, their J. Martin, review for a Canon battery)
management has come to occupy a more prominent role in The Canon battery is essential to have. Knock-off brands don’t
marketing practice. Likewise, consideration of the review char- last a quarter as long as the Canon branded batteries. (“tac cat,”
acteristics exerting the most impact has come to occupy a more review for a Canon battery)
prominent role in marketing theory. Positive reviews generally
boost sales while negative reviews hurt sales (Chevalier and
Mayzlin 2006; see also Ho-Dac, Carson, and Moore 2013), and Taly Reich is an Associate Professor in Marketing, Yale School of Management,
Yale University (email: taly.reich@yale.edu). Sam J. Maglio is an Associate
firms certainly opt to feature positive over negative reviews. Professor in Marketing, Department of Management, and Department of
Packard and Berger (2017) presented evidence going one step Psychology, University of Toronto Scarborough, and Rotman School of
further, examining not simply whether the review was positive Management (email: sam.maglio@utoronto.ca).
Reich and Maglio 53

Both of these reviews are positive, but which would—and stable, permanent cause; Dweck and Leggett 1988; Hong,
which should—Canon feature: Richard, who previously pur- Chiu, and Dweck 1995; Peterson et al. 1982). Rather, we the-
chased a product that turned out to be a mistake, or “tac cat,” orize that a reviewer’s admitted purchase mistake signals a
whose purchase experience was not marred by a mistake? temporary lack of knowledge, which is an unstable cause that
The present investigation develops a theoretical model in can change over time (Dweck and Leggett 1988; Hong, Chiu,
proposing that admitting to having made a previous mistake and Dweck 1995; Kupor, Reich, and Laurin 2018; Peterson
acts as a powerful cue through which potential purchasers infer et al. 1982). Provided that a lack of knowledge is temporary
that the reviewer has gained significant expertise about a focal and, thus, fixable, we further propose that reviewers who admit
product domain, which in turn increases the potential purcha- to having made a mistake will prove especially likely to be seen
sers’ likelihood of purchasing the product that the reviewer as having rectified this lack of knowledge (i.e., to have gained
recommends. In keeping with academic and applied norms, expertise since the purchase mistake; we conceptualize exper-
we refer to the communication under consideration as a tise as a subcomponent of the broader construct of consumer
“review” and the communicator as a “reviewer.” However, knowledge in keeping with Alba and Hutchinson [1987]).
defined formally, reviews need only describe the reviewer’s These predictions derive from two lines of reasoning. First,
experience, whereas “recommendations” advise on what to to echo and advance a point made previously, all mistake
do. Because we want to better understand how the content of admissions require a mistake to have occurred, but not all
communication shapes its ability to persuade, the communica- mistakes that are made result in the admission of a mistake.
tions in our experiments will be presented as reviews yet In other words, only a subset of mistaken purchasers will ever
worded as recommendations that prescribe a certain course of admit to their mistakes, and even fewer mistaken purchasers
action to allow us to measure the degree to which our partici- will admit to their mistakes in a public forum (like online
pants are persuaded (cf. Yeomans 2019). reviews). We propose that admitting to having made a mistake
inherently conveys that the mistaken reviewer has gained new
expertise. In the case of product purchases, the mistaken
Admitting Mistakes: Helpful or Harmful for
reviewer, in admitting the mistake, must now know not only
Persuasion? that the original product has fallen short of initial expectations
Drawing from previous conceptual work, we define a purchase (forming the basis of a negative assessment of the product) but
mistake as a self-identified suboptimal decision whereby peo- also that a different course of action (purchasing a different
ple purchase a product that subsequently fails to meet a thresh- product) would have proven better (for other opportunities to
old level of expected performance (Huei-Chen 2007; Laband learn from mistakes and signal that learning to others in online
1986; Murphy and Enis 1986). Accordingly, we are not con- reviews, see the “General Discussion” section). On top of this
cerned with mistakes of execution, whereby someone pur- (which applies even in admitting a mistake to oneself), we
chases a product unintentionally or by accident. Rather, as an reason that the public admission of having made a mistake
example of a mistake in the form of a suboptimal decision, a signals that mistaken reviewers are especially confident in their
consumer might purchase a speaker system only to find that the current assessment (e.g., product review), resulting from new
speakers do not perform as well as anticipated. Does this pur- expertise gained, insofar as they are willing to engage in the
chase mistake—specifically, as admitted by the mistaken pur- costly behavior of conceding something negative about them-
chaser—alter the willingness of other potential purchasers to selves. Conversely, when reviewers acknowledge switching
heed the advice of this mistaken purchaser? between brands without admitting a mistake, that reviewer’s
One possibility is that admitting to having made a mistake reason for switching remains more vague: admitting a mistake
undermines other people’s inclination to follow any advice signals a gaining of expertise, whereas no such admission could
from the mistaken reviewer. After all, admitting to a mistake result from any number of factors (e.g., promotions, stockouts,
necessitates the making of a mistake, and mistakes may signal variety seeking; Sun, Neslin, and Srinivasan 2003; Van Trijp,
incompetence: The mistaken reviewer was not sufficiently Hoyer, and Inman 1996; Walter and Grabner 1975).
thoughtful, intelligent, and/or knowledgeable to adequately Second, we propose that the impact on others of a reviewer
assess the relevant product specifications and make a good admitting to a mistake also results from the experience of those
purchase decision (Angelis et al. 2012; East, Hammond, and others (here, potential purchasers reading online reviews) in
Wright 2007; Wojnicki and Godes 2017). People often judge making mistakes of their own. Everyone makes mistakes, and
others who make mistakes as incompetent and punish them a common response is to exert additional effort to learn from
(e.g., Chesney and Su 2010; Gawande 1999; Kunda 1999; Par- the mistake to avoid making a similar mistake again in the
ker and Lawton 2003), so awareness of a reviewer’s previous future (Argyris and Schon 1978; Carver and Scheier 2012;
purchase mistake might predict an outcome resulting from an Locke et al. 1981; Podsakoff and Farh 1989). Indeed, making
inference of incompetence and akin to punishment: refusing to a mistake (vs. enjoying success) prompts greater subsequent
take the reviewer’s advice. effort in the mistaken domain through the use of self-regulation
Another possibility, which forms the foundation of our the- (Carver and Scheier 2012): the mistake signals a discrepancy
oretical model, is that people do not necessarily attribute oth- between a desired state (e.g., owning a good product) and the
ers’ admission of a purchase mistake to incompetence (i.e., a current state (e.g., having made a bad purchase), which triggers
54 Journal of Marketing 84(1)

attempts to reduce the discrepancy and attain the desired goal H4: Consumers are more likely to purchase a product
(Campion and Lord 1982; Podsakoff and Farh 1989). People recommended by reviewers who admit to having made
experience this effect consciously: they are often clearly aware a previous purchase mistake (vs. reviewers who mention
that making a mistake causes them to exert more effort in having made a successful purchase) in the same domain,
similar situations in the future (Campion and Lord 1982). Just whereas consumers are no more likely to purchase a prod-
as people themselves exert greater effort following a mistake uct recommended by reviewers who admit to having
(vs. a success), they may believe that others exert greater effort made a previous purchase mistake (vs. reviewers who
after their mistakes as well. Because people assume that greater mention having made a successful purchase) in a different
effort produces better outcomes (Kruger et al. 2004), we for- domain.
mally predict the following:

H1: Consumers infer that reviewers who admit to having Research Overview
made a previous purchase mistake (vs. reviewers who Study 1 tests our predictions that people infer mistaken
mention having made a successful purchase) have more reviewers to have more expertise in the product domain in
expertise in the mistaken product domain. which the mistake was made (H1), that purchase advice is more
We also predict that these inferences shape the persuasive likely to be accepted from a mistaken reviewer (H2), and that
impact of reviews. This is because the perception of a the former accounts for the latter (H3). Study 1 tests these
reviewer’s expertise is a primary determinant of whether peo- predictions using an incentive-compatible design, and Study
ple purchase the products that a reviewer recommends: people 2 tests these same predictions (H1–H3) but by assessing inferred
are more likely to follow the recommendations of those who expertise with a different measure in the interest of providing a
appear to have more expertise about the domain in which they robustness check. Study 3 manipulates not only admission of a
are making a recommendation (Packard and Berger 2017; Price mistake but also the domain alignment of the mistake, intro-
and Stone 2004; Wen, Tan, and Chang 2009; Woodside and ducing a key moderator to provide evidence for the boundary
Davenport 1974). At this downstream level, we predict the condition articulated in H4. Study 4 introduces several mod-
following: ifications to the general design of Studies 1–3 to provide
evidence for a robust effect relating mention of making a
H2: Consumers are more likely to purchase a product mistake to acceptance of purchase advice (H2) for a real pur-
recommended by reviewers who admit to having made chase decision. Finally, Study 5 examines mistaken
a previous purchase mistake (vs. reviewers who mention reviewers’ persuasive impact in the field by examining real
having made a successful purchase). reviews from a popular website (Sephora). Thus, these data
suggest that the persuasive power of mistakes is sufficiently
H3: Consumer inferences of expertise for reviewers who
robust to emerge in the noisy real world.
admit to having made a previous purchase mistake (vs.
The inclusion of orthogonal experimental factors in these
reviewers who mention having made a successful pur-
designs indirectly speaks to several alternative explanations
chase) mediate the relationship between admission of a
that we discuss in turn, and a posttest of Study 3 directly mea-
mistake and likelihood of purchasing the recommended
sures other inferences consumers might make of reviewers who
product.
admit mistakes, finding that they do not account for the effect
Our hypothesis development has centered on inferences of observed throughout the present investigation. We identify sev-
expertise, but we do not propose that any admission of having eral such possibilities for these findings. First, a positive review
made a previous mistake always enhances perceived expertise for a focal product that includes admission of a mistake
in a way that should increase advice taking. Rather, consumers acknowledges the existence of both well-performing and
should be able to discern whether a review in which a mistake underperforming brands. This may result in a brand compari-
was admitted presents compelling evidence that the reviewer son effect, whereby any mention of poor performance for one
has learned from that mistake. As an implication of this pre- brand makes the other, well-performing brand look better (con-
dicted sensitivity on the part of consumers, we propose—as a sistent with the effectiveness of comparative advertising; Gre-
boundary condition to our main effect—that consumers should wal et al. 1997). A second, related explanation might hold not
be more likely to follow the purchase-related advice of for perception of the two brands but for perception of the
reviewers who admit to having made a previous mistake only reviewer: in mentioning the two brand performances, observers
when the mistake conveys that the reviewer has gained exper- might infer that the mistaken reviewer has more carefully con-
tise since making the mistake. Specifically, we theorize that the sidered both the positives and the negatives of the focal alter-
expertise only should be seen as strengthened in the domain in natives. Third, a review that includes the admission of a
which the reviewer made the mistake. Accordingly, potential mistake necessarily includes both positive information (a
purchasers should heed the mistaken reviewer’s advice for favorable review for the focal product) and negative informa-
products in that domain but, provided the expertise does not tion (that the previous purchase was a mistake), in contrast to
transfer readily across domains, not for products in any other the unilateral positivity of the successful purchase conditions.
domain. Formally, we predict the following: As a result, perhaps exposure to any negative information in a
Reich and Maglio 55

review orients consumers to potential losses, looming larger would be entered into a lottery for a prize. Participants further
than potential gains and prompting them to accept more readily read that if they won the lottery, they would receive a pair of
the advice in a review (e.g., as the result of risk aversion: headphones and that they would choose which one of two sets
Kahneman and Tversky 1979). Should this be the case, then of headphones they preferred to receive as their prize. All
mention of any mistake might strengthen the tendency to participants then viewed information about the two sets of
accept the advice in a review. These alternative explanations headphones, which were called Orbin and Raymour (see Web
would predict that consumers should still be persuaded by a Appendix A). They also read a consumer review that was
reviewer who first makes a mistake in one product domain and (ostensibly) the most recently submitted review for these head-
then makes a successful product purchase in a different phones. Specifically, participants saw that the most recently
domain. Instead, we predict and test in Study 3 that admission submitted review was written by a consumer named Sam. Par-
of a mistake will change only inferred expertise in the same, ticipants were randomly assigned to one of two conditions: a
focal product domain, rendering moot advice regarding other mistaken reviewer condition or a successful reviewer condi-
product domains (H4). Separately, reviewers who admit mis- tion. In both conditions, Sam’s review recommended the Orbin
takes might be perceived differently on multiple different headphones (purchased most recently) and also described a
inferred characteristics by observers, including specific person- previous purchase of headphones, providing his experience
ality traits and as having global expertise that extends across with both sets of headphones in terms of general evaluation
different product domains. Accordingly, a posttest of Study 3 as well as performance on the same particular attribute (a sen-
measures several such potential inferences, finding no connec- sor). In the mistaken reviewer condition, Sam noted that this
tion between them and the admission of a mistake. previous purchase was a mistake:
While we designed Study 3 with the goal of speaking against
several alternative psychological explanations for our effect, A couple years ago when I was searching for the last pair of
we designed Study 4 with the goal of speaking to practitioners headphones that I bought, I ended up buying the Nidec VIA
interested in the breadth by which our effect might be applied. headphones, and that was a mistake—it turned out that the
For this reason, Study 4 departs from Studies 1–3 in subtle but headphones had a bad type of sensor and therefore did not work
meaningful ways. First, it presents not a single review but a well. I recently decided to upgrade my headphones to a newer
set of ten reviews in which we either did or did not embed a model, and I considered both the Orbin and Raymour head-
single review in which the reviewer mentions making a pre- phones. I chose the Orbin headphones. I’ve had them for a
vious purchase mistake. Second, rather than reviews for ficti- month, and they are good—they have great features, including
tious brands or brands with which participants ostensibly have a good type of sensor, and they work well. I would recommend
little familiarity, Study 4 presents reviews for a known brand. them. (Sam K.)
Using a known brand facilitated our third change: having
participants make a consequential choice. To compliment the Conversely, in the successful reviewer condition, Sam noted
consequential setup in Study 1 (in which one participant that this previous decision was successful:
received a chosen outcome), all participants in Study 4
received the outcome they choose. Thus, Study 4 attests to A couple years ago when I was searching for the last pair of
the robustness of our effect in tandem with the large-scale headphones that I bought, I ended up buying the Nidec VIA
data-mining approach adopted in Study 5. headphones, and that was a good choice—it turned out that the
headphones had a good type of sensor and therefore worked
well. I recently decided to upgrade my headphones to a newer
Study 1 model, and I considered both the Orbin and Raymour head-
In an incentive-compatible context, Study 1 tests H1–H3: con- phones. I chose the Orbin headphones. I’ve had them for a
sumers infer that a reviewer has more expertise about a product month, and they are good—they have great features, including
category if the reviewer admits to previously making a mistake a good type of sensor, and they work well. I would recommend
them. (Sam K.)
in purchasing a product from that category than if the reviewer
does not, and this inference of expertise accounts for the greater
tendency of consumers to choose in line with this reviewer’s Next, participants were asked whether they preferred to
advice. receive the Orbin or the Raymour headphones if they won the
lottery. The instructions emphasized that this decision was real.
Participants entered their decision by selecting a radio button
Method that was labeled either with “Orbin” or “Raymour.” They also
One hundred sixty participants (mean age ¼ 35 years; 39% rated how much they perceived that the reviewer had learned
male) from a large East Coast U.S. university participated in about how to choose good headphones (1 ¼ “Not much at all,”
a laboratory study in exchange for course credit. This study was and 7 ¼ “A lot”). At the end of the study, participants were
run as part of a session containing unrelated surveys from debriefed (i.e., they were informed that the lottery was real but
different researchers. All participants read that, as additional was for a pair of Sony headphones rather than for the brands
compensation for their participation in the lab session, they that they read about in the study).
56 Journal of Marketing 84(1)

a purchase mistake through an inference that those reviewers


Perceived have learned more. As we have noted, we believe that an
Learning
inference of learning implies the gaining of new consumer
.99*** .06** knowledge in the form of expertise, though Study 1 assessed
only the former. Thus, in Study 2, we test the same hypotheses
as Study 1 (H1–H3) using a modified empirical approach. Spe-
Mistaken vs. Consumer cifically, we utilize a hypothetical scenario design to comple-
Successful Reviewer (.08) .14* Choice ment the incentive-compatible design of Study 1, and we
measure our proposed mediator by asking whether perceivers
Figure 1. Mediation model in Study 1. judge mistaken reviewers to have more knowledge in the rel-
Notes: The path coefficients are unstandardized betas. Values in parentheses evant domain compared with successful reviewers.
indicate the effect of condition on the dependent variable after controlling for
the mediator. 95% CI for the indirect effect ¼ [.1614, .8766].
*p < .05. Method
**p < .01.
***p < .001. Eighty participants (mean age ¼ 34 years; 63% male) in an
online participant pool participated in a study in exchange for
monetary payment. All participants imagined that they lived in
Results and Discussion Seattle and were looking for a local florist. They further ima-
As we predicted, participants more often chose the recom- gined that they had narrowed their choices down to two local
mended headphones when they were recommended by a mis- artisan floral shops that served only the local Seattle area—
taken reviewer (93.1%) than when they were recommended by FlowersNow and FreshBlooms. Participants viewed informa-
a successful reviewer (79.5%; w2(d.f. ¼ 1, N ¼ 160) ¼ 5.87, p tion about the two floral shops (Web Appendix B) and ima-
¼ .015; Cohen’s d ¼ .390). Furthermore, participants perceived gined that they looked at the websites of the two florists to help
that the mistaken reviewer learned more about how to choose them make their decision. On the FreshBlooms website, parti-
good headphones (M ¼ 5.96, SD ¼ 1.01) than the successful cipants saw that there was a review written by a consumer
reviewer (M ¼ 4.97, SD ¼ 1.39; t(158) ¼ 5.07, p < .001, named Sam. In both conditions, Sam’s review recommended
Cohen’s d ¼ .811). FreshBlooms and also described a previous choice he had made
We hypothesized that the mistaken (vs. successful) reviewer between two florists when he lived in a different state. In the
more strongly influenced participants’ real choices because mistaken reviewer condition, Sam noted that this previous
participants inferred that the mistaken reviewer gained more decision was a mistake:
expertise about how to choose good headphones. Consistent
with this prediction, a mediation analysis with 5,000 bootstraps I just moved to Seattle from Boston. When I was in Boston, I
revealed that perceived learning mediated the effect of condi- decided to get flowers once from a local florist in Boston (which
tion on participants’ real choices (95% confidence interval [CI] sells flowers only in the Boston area). That was a mistake—the
for the indirect effect ¼ [.1614, .8766]; see Figure 1). florist I chose was not a good florist. Now that I’ve moved to
In summary, Study 1 documents the persuasive power of Seattle, I wanted to get flowers for my housewarming party. My
mistakes in an incentive-compatible context. Moreover, Study final choice came down to flowers from FlowersNow or Fresh-
1 suggests that this power arises because people infer that a Blooms. After looking into both options, it was clear to me that
FreshBlooms is the better florist. I decided to get flowers from
purchaser who makes a mistake subsequently gains more
FreshBlooms, and that was a great choice. I recommend Fresh-
expertise—assessed here in the form of learning—about how
Blooms! (Sam K.)
to identify a good product in that domain than a purchaser
who originally chose successfully. Here, the mistaken
Conversely, in the successful reviewer condition, Sam noted
reviewer identified the previous purchase as a mistake on the
that this previous decision was successful:
basis of its performance on one particular attribute (a sensor),
creating potential confounds (e.g., reviewer evaluation of this I just moved to Seattle from Boston. When I was in Boston, I
highly technical attribute leading to observer inferences of decided to get flowers once from a local florist in Boston (which
preexisting expertise) that we address in our subsequent stud- sells flowers only in the Boston area). That was a good choice—the
ies by having the mistaken reviewers concede only that they florist I chose was a good florist. Now that I’ve moved to Seattle, I
made previous purchase mistakes without detailing the rea- wanted to get flowers for my housewarming party. My final choice
sons for this conclusion. came down to flowers from FlowersNow or FreshBlooms. After
looking into both options, it was clear to me that FreshBlooms is
the better florist. I decided to get flowers from FreshBlooms, and
Study 2
that was a great choice. I recommend FreshBlooms! (Sam K.)
Study 1 provides, in a consequential choice setting, evidence
consistent with the prediction that consumers are more per- Next, participants reported whether they would choose to
suaded by reviewers who have (vs. have not) previously made buy flowers from FreshBlooms or FlowersNow. They also
Reich and Maglio 57

reviewer has significant expertise (H3). As a result, any infor-


Perceived mation that undermines the tendency for observers to infer that
Knowledge
the reviewer has this expertise should, in turn, undermine the
.84** .14*** extent to which the review is persuasive. Thus, to provide
convergent evidence, Study 3 utilizes a moderation-of-
process design (Spencer, Zanna, and Fong 2005) that manip-
Mistaken vs. Consumer ulates not only the admission of a mistake but also an
Successful Reviewer (.18) .29** Choice additional experimental factor designed to compromise the
inferred expertise of the reviewer, which also addresses
Figure 2. Mediation model in Study 2. potential alternative accounts. Specifically, if the persuasive
Notes: The path coefficients are unstandardized betas. Values in parentheses power of mistakes is due to the belief that reviewers’ purchase
indicate the effect of condition on the dependent variable after controlling for mistakes signal that the reviewers have gained more expertise
the mediator. 95% CI for the indirect effect ¼ [.0545, .2834]. about the product domain, then mistaken reviewers’ recom-
*p < .05.
**p < .01.
mendations will be more persuasive only when the reviewers’
***p < .001. purchase mistakes occur in the same product category as their
focal review (H4). We tested this prediction in Study 3 by
presenting participants with a review from either a mistaken
completed a two-item index of their perceptions of the
or successful reviewer, in keeping with the designs of Studies
reviewer’s knowledge. Specifically, they indicated how much
1 and 2, but departing from those studies in that the previous
knowledge the reviewer now had about how to choose a good
purchase was in either the same or a different product cate-
florist and how much knowledge the reviewer had about
gory as the review.
whether to buy flowers at FreshBlooms or FlowersNow. Parti-
cipants responded to each question on separate seven-point
scales (1 ¼ “Not a lot,” and 7 ¼ “A lot”). We combined the
items into an index of perceived knowledge (r ¼ .53, p < .001). Method
We predicted that the mistaken (vs. successful) reviewer
Two hundred ninety-nine participants (mean age ¼ 32 years;
would more strongly influence participants’ choices because
49% male) from Amazon’s Mechanical Turk (MTurk) com-
participants would perceive the mistaken reviewer as more
pleted an online study in exchange for monetary payment.
knowledgeable.
The participants’ task was to decide which one of two in-
ceiling speaker systems they would prefer to purchase: the
Results and Discussion Mikana XPI in-ceiling speaker system or the Rokana SX2 in-
As we predicted, participants more often chose the recom- ceiling speaker system. Participants viewed the specifications
mended florist when it was recommended by a mistaken of the two speaker systems (which were identical to the spec-
reviewer (90.9%) than when it was recommended by a success- ifications of the headphones described in Study 1, as the
ful reviewer (61.7%; w2(d.f. ¼ 1, N ¼ 80) ¼ 8.54, p ¼ .003, specifications could reasonably apply to both product cate-
Cohen’s d ¼.691). Furthermore, participants perceived that the gories; see Web Appendix A) and read two reviews written
mistaken reviewer had more knowledge (M ¼ 4.68, SD ¼ 1.27) by a consumer named Taylor. Both reviews were ostensibly
than the successful reviewer (M ¼ 3.84, SD ¼ 1.19; t(78) ¼ on a website that featured consumer reviews for different
3.03, p ¼ .003, Cohen’s d ¼ .683). To test the mediating role of electronics. The focal review, which was the same in all
perceived knowledge in determining the effect of condition on conditions, noted that Taylor purchased a Mikana XPI
choice, we conducted a mediation analysis with 5,000 boot- speaker system and recommended it. However, the content
straps. As we hypothesized, perceived knowledge mediated the of the nonfocal review differed by condition. Specifically,
effect of condition on choice (95% CI for the indirect effect ¼ participants were randomly assigned to read that Taylor had
[.0545, .2834]; see Figure 2). previously purchased a bookshelf speaker (i.e., a product in
Along with Study 1, Study 2 provides evidence for the per- the same domain as the focal product) or a printer (i.e., a
suasive power of mistakes. Moreover, in Study 2, we establish product in a different domain from the focal product) and
that this power arises because consumers infer that a reviewer that this decision was either a mistake or a success (see Web
who admits to a mistake has more knowledge than an equiva- Appendix C). These manipulations produced a 2 (nonfocal
lent reviewer who originally chose successfully. Having docu- product type: speakers vs. printer)  2 (nonfocal review type:
mented a robust effect in Study 2, in Study 3 we probe a mistake vs. success) design.
potential moderator of the persuasive power of mistakes. After participants viewed Taylor’s reviews (both reviews
were explicitly noted as written by Taylor), they decided
whether they would purchase the Mikana XPI or the Rokana
Study 3 SX2 speaker system. Participants entered their decision by
Our theoretical model proposes that the persuasive power of selecting a radio button that was labeled either with “Mikana
reviews featuring mistakes comes from an inference that the XPI” or “Rokana SX2.”
58 Journal of Marketing 84(1)

Table 1. Descriptive Statistics in the Posttest of Study 3.

Speaker Printer
DV Speaker Mistake Success Printer Mistake Success Analysis (Interaction)

Discerning 4.21 (1.49) 4.13 (1.40) 4.53 (1.31) 4.18 (1.40) F(1, 346) ¼ .83, p ¼ .362
Integrity 4.76 (1.18) 4.48 (.97) 4.78 (.90) 4.85 (1.10) F(1, 346) ¼ 2.52, p ¼ .114
Liking 4.27 (1.34) 4.23 (1.33) 4.52 (1.19) 4.55 (1.28) F(1, 346) ¼ .06, p ¼ .810
Similar 5.56 (1.45) 5.76 (1.16) 6.17 (.95) 6.15 (1.02) F(1, 346) ¼ .73, p ¼ .393
Surprise 4.66 (1.22) 4.60 (1.27) 4.89 (1.17) 4.80 (1.23) F(1, 346) ¼ .01, p ¼ .930
Expertise Index 4.64 (1.13) 4.31 (.99) 4.59 (1.09) 4.74 (1.01) F(1, 346) ¼ 4.43, p ¼ .036
Notes: Statistics in parentheses are standard deviations.

Results We recruited a separate sample of MTurk participants (N ¼


350; mean age ¼ 38 years; 50% male), who were randomly
We conducted a binary logistic regression using nonfocal
assigned to view the same information presented to participants
review type (mistake vs. success), nonfocal product type
in the main study. Posttest participants then completed mea-
(speakers vs. printer), and their interaction to predict partici-
sures, presented in a random order, assessing their perceptions
pants’ choices. The regression revealed an interaction on
of the extent to which Taylor was discerning and had high
choice (b ¼ 2.68, z ¼ 2.67, p ¼ .008). In a conceptual replica-
integrity, the extent to which they liked Taylor and were similar
tion of the previous studies, when the reviewer had made a
to Taylor, and the extent to which Taylor’s reviews were sur-
previous speaker purchase, participants who read that the pre-
prising. In addition, participants completed a two-item index of
vious purchase was a mistake chose the recommended speaker
their perceptions of Taylor’s expertise (adapted from Packard
more often (97.4%) than participants who read about a success-
and Berger 2017). Specifically, they indicated how much of an
ful purchase (87.0%; b ¼ 1.72, z ¼ 2.17, p ¼ .030, odds ratio ¼
expert they thought Taylor was about speakers and how knowl-
5.597). Conversely, when the reviewer had made a previous
edgeable they thought Taylor was about speakers. Similar to
printer purchase, we found no such effect (mistake ¼ 86.5%,
Packard and Berger (2017), we combined these items into an
success ¼ 94.4%; b ¼ .96, z ¼ 1.56, p ¼ .119).
index of perceived expertise that correlated at a level (r ¼ .66,
p < .001) commensurate with their original work (r ¼ .53).
Participants indicated their responses on separate seven-point
Posttest scales (1¼ “Not at all,” and 7 ¼ “Very much”).
As we expected, the interaction between nonfocal product
If a mistake’s persuasive power arises because admitting a
type and nonfocal review type was significant for perceived
mistake signals some broadly positive character trait (e.g., the
expertise (F(1, 346) ¼ 4.43, p ¼ .036). When the reviewer had
integrity to admit one’s mistakes publicly), then mistaken
made a previous speaker purchase, the mistaken reviewer was
reviewers should be more persuasive regardless of the domain
perceived as more expert (M ¼ 4.64, SD ¼ 1.13) compared
of their mistake. Thus, we conducted a posttest to establish
with the successful reviewer (M ¼ 4.31, SD ¼ .99; Fisher’s
whether mentioning a mistake—either in the same domain or
least significant difference: p ¼ .043; Cohen’s d ¼ .311). In
a different domain as the focal product—alters how consumers
contrast, when the reviewer had made a previous printer pur-
perceive the mistaken reviewer. The design of the posttest
chase, there were no differences in perceived expertise between
mirrored that of the main study, save for a switch from measur-
the mistaken reviewer (M ¼ 4.59, SD ¼ 1.09) and the success-
ing choice to measuring several such potential perceptions.
ful reviewer (M ¼ 4.74, SD ¼ 1.01; F < 1). Participants’
Specifically, the posttest began with participants’ assessment
perceptions of the extent to which Taylor was discerning, had
of the extent to which a mistaken reviewer is discerning, has
integrity, was likeable, and was similar to themselves, as well
integrity, is likeable, and is similar to the self. The posttest
as the extent to which their mistake was surprising did not
also measured whether the mistake was surprising, because
differ as a function of product type and review type (Fs(1,
surprise can orient attention to a focal piece of information
346) < 2.52, ps > .114; see Table 1).
and increase persuasion as a result (Baker and Petty 1994).
Finally, in the interest of presenting evidence not only against
these alternative explanations but also in support of our pro-
posed mechanism, participants rated the perceived expertise Discussion
of the mistaken reviewer to conceptually replicate the media- These results are inconsistent with the possibility that a mis-
tion results from Studies 1 and 2. We examined these issues take’s persuasive power arises because they signal some broad
using a posttest rather than in the main study to avoid potential character trait that enhances discernment, integrity, likeability,
demand effects from asking participants about both their pur- or perceived similarity to the observer. Theoretically, such
chase intentions and their perceptions of the mistaken character traits should have been signaled just as well by a
reviewer in the same study. printer-related mistake as by a speaker-related mistake, in
Reich and Maglio 59

which case we would not have found the predicted interaction. were told that they would be asked to read several reviews of
Moreover, Study 3’s posttest confirmed that Taylor was per- spearmint mints made by a brand of which we presumed parti-
ceived equivalently on each of these dimensions regardless of cipants to have at least some knowledge (Altoids) and then
whether he made a prior mistake purchasing a printer or asked to make a choice. Participants were randomly assigned
speakers. Study 3, however, does not speak to one remaining to one of two conditions: a mistake condition or a no-mistake
potential inference: that the mistaken reviewer had high condition. All participants viewed a total of ten reviews that
expertise from the outset instead of gaining expertise. were presented in a random order. Nine of these reviews were
Although at face value this possibility might seem inconsis- identical between the experimental conditions and were taken
tent with making a mistake in the first place, we conducted a from actual Altoids spearmint mints reviews posted on Ama-
supplementary study (reported in Web Appendix D) to zon (including star rating, title of review, and review text; for
demonstrate that the persuasive power of mistakes arises not all reviews, see Web Appendix E). We varied the content of
because people assume that a mistaken reviewer had a lot of the remaining review to either include a reference to the
knowledge to begin with to discern their mistake (and not reviewer having made a previous mistake (mistake condition)
because of any other variable that consumers may assume or not (no-mistake condition). After reading the reviews, par-
plays into reviewers’ ability or willingness to acknowledge ticipants were told that as additional compensation for the
a mistake, or because of some other difference in the content survey session, they could choose to receive either one pack
of mistake- vs. success-based reviews), but rather because of Altoids spearmint mints or one additional dollar added to
they assume that a mistaken reviewer has acquired more their payment. Participants entered their decision by selecting
expertise as a result of their mistake. a radio button that was labeled either with “Altoids Spearmint
Mints” or “One Additional Dollar,” and the researcher then
provided the participant with their chosen form of additional
Study 4 compensation.
The first three studies identified the psychological foundation
underlying why mentioning a mistake causes consumers to Results and Discussion
place more credence in the advice of those reviewers. In the
pursuit of this objective, the first three studies prioritized inter- A chi-square analysis revealed that participants reading a set of
nal validity over external validity; Study 4 shifts its balance to reviews that contained one review in which the reviewer men-
consider the impact of mistaken reviewers in more ecologically tioned making a mistake were more likely to choose the Altoids
valid contexts. While considering a purchase, consumers reg- spearmint mints over additional monetary compensation
ularly read not only a single review in isolation but, rather, (34.9%) than were participants for whom the provided set of
multiple reviews to form an overall conclusion. To capture this reviews did not contain a review mentioning a mistake (22.0%;
reality, participants see not one but ten reviews for a focal w2(d.f. ¼ 1, N ¼ 249) ¼ 5.14, p ¼ .023; Cohen’s d ¼ .290).
product in Study 4; two experimental conditions vary whether Thus, as we predicted, even when the review mentioning a
one review, embedded within that set of ten, mentions a prior mistake has only a minority presence (i.e., comprises one
purchase mistake. Furthermore, would the effect of mentioning review out of ten), it still exerts an effect powerful enough to
mistakes extend from the fictitious or generic brands used in change consumer choice. Notably, the behavior under consid-
our previous studies to established, known brands presumably eration here closely reflects real consumer decision making, as
higher in brand equity? Study 4 considers this robustness issue our participants learned about a widely known brand and sub-
by using a real brand (Altoids mints) as the focal product under sequently made a consequential choice (between receiving the
consideration. Finally, as this study aims to provide the clearest branded product or receiving additional money). These results
point of direct application, it requires all participants to make a attest to the strength and robustness of our effect, which we
real purchase choice. Whereas the choice in Study 1 was extend in Study 5 to a different naturalistic context.
incentive-compatible insofar as one participant would ulti-
mately receive the chosen outcome, incentive compatibility is
strengthened in Study 4, which asks all participants to make a Study 5
real purchase decision. We predicted that mention of a mistake
would still prove powerful under this more naturalistic set of In our final study, we test the external validity of our findings
conditions. by examining whether the persuasive power of mistakes is
sufficiently robust to emerge in the noisy real world. To that
end, we examine consumer reviews on Sephora’s online retail
Method platform. Conveniently, the Sephora review platform has a
Two hundred forty-nine participants (mean age ¼ 35 years; feature whereby readers can rate whether reviews are helpful,
43% male) from a large U.S. East Coast university participated which is indicative of whether they are persuasive (see Baek,
in a session of laboratory studies in exchange for monetary Ahn, and Choi 2012; Mudambi and Schuff 2010; see also the
payment. This study was run as part of a session containing pilot test in the “Results and Discussion” subsection of this
unrelated surveys from different researchers. All participants study). We predict that reviews referencing a purchase mistake
60 Journal of Marketing 84(1)

will be linked to consumers finding the review more helpful (as analysis revealed that most people (72.2%) reported that they
measured by reader-provided ratings of helpfulness). rate a review as helpful when it is persuasive (w2 (d.f. ¼ 1) ¼
14.22, p < .001). Thus, consistent with prior literature (Baek,
Method Ahn, and Choi 2012; Mudambi and Schuff 2010), reviews’
helpfulness ratings serve as a proxy for their persuasive power.
The Sephora website includes six product categories (makeup,
skincare, hair, tools and brushes, fragrance, and bath and body). Primary analyses. We computed a measure of each review’s
The category to be scraped (hair) was randomly chosen and, persuasive power by subtracting the number of unhelpful votes
after determining 40 products as the maximum number able to from helpful votes associated with each review and dividing
be scraped within a reasonable time frame, 40 haircare products that by the total number of votes (helpful  unhelpful)/(helpful
were randomly chosen. Within that subset, we scraped all þ unhelpful). This measure served as our dependent variable,
reviews starting in August 2017 until the time of scraping and it ranged from 1 to 1 (M ¼ .078, SD ¼ .329; Mhelpful ¼
(December 24, 2018). For the entire resulting set of 5,727 3.672, SD ¼ 16.621; Munhelpful ¼ .774, SD ¼ 3.285). As a first
reviews, we used a series of indicator variables to tag whether step, we regressed this helpfulness measure on whether the
each review mentioned a previous purchase mistake. Specifi- review referenced a mistake (1 ¼ no, 1 ¼ yes). As we pre-
cally, the (case-insensitive) indicators were: mistak: the string dicted, the regression revealed that reviews referencing a mis-
“mistak” is in the review; mistook: the string “mistook” is in take (vs. those not referencing a mistake) were deemed more
the review; my_bad: the phrase “my bad” or “my error” is in helpful (b ¼ .076, SE ¼ .010, p < .001). Next, we regressed the
the review; I_wrong: the word “I” is within 35 characters of a helpfulness measure on whether the review referenced a mis-
word starting with “wrong” (without a period, question mark, take (1 ¼ no, 1 ¼ yes) and the following control measures (all
or exclamation mark in between, which are proxies for sen- continuous control variables were mean-centered): review
tence divisions); my_fault: the word “my” is within 35 charac- length (number of words), valence of review (1 ¼ negative,
ters of the word “fault” (again, without a period, question mark, 0 ¼ both positive and negative, 1 ¼ positive; coded by two
or exclamation mark in between); and our_fault: the word independent judges with high interjudge reliability: r ¼ .93),
“our” is within 35 characters of the word “fault” (again, with- star rating (1–5), explicit recommendation (1 ¼ no, 1 ¼ yes;
out a period, question mark, or exclamation mark in between). feature included on the Sephora website), loyalty program
This tagging led to the identification of 502 reviews referen- membership tier (with higher numbers indicating more dollars
cing a prior mistake. Two independent judges (interjudge relia- spent at Sephora in a calendar year; 1 ¼ “Insider,” 2 ¼ “VIB,”
bility: r ¼ .92) reviewed these 502 reviews, tagging 86% of and 3 ¼ “Rouge”), reviewer expertise (with higher numbers
them to be about mistakes in choice.1 To equate sample sizes, indicating more reviews posted to the Sephora site; 1 ¼
we then randomly selected 502 reviews that did not reference a “Rookie,” 2 ¼ “Rising Star,” 3 ¼ “Go Getter,” and 4 ¼
mistake from the remaining scraped data set, resulting in a data “Boss”), number of images uploaded with the review, explicit
set of 1,004 reviews. mention of another brand in the review (1 ¼ no, 1 ¼ yes;
coded by two independent judges with high interjudge relia-
Results and Discussion bility, r ¼ .94), and date of review (calculated as number of
days since December 30, 1899 on the Gregorian calendar). The
Pilot test. Previous research has suggested that helpfulness votes regression revealed a significant effect of review length on
are a proxy for persuasiveness (Baek, Ahn, and Choi 2012; helpfulness, such that longer reviews were found to be more
Mudambi and Schuff 2010). We conducted our own pilot test helpful (b ¼ .001, SE < .000, p < .001). None of the other
to further verify this conclusion. That is, we tested whether control variables had a significant effect on helpfulness (ps >
consumers rate a review as helpful when it guides their pur- .304). Most relevant to our focal theorizing, the regression also
chase decision. To that end, we recruited 72 participants from revealed that reviews referencing a mistake (vs. those not refer-
MTurk who reported that they had previously rated an Amazon encing a mistake) were deemed more helpful in the full analysis
review as helpful or unhelpful. We then asked them to describe controlling for review length, valence of review, star rating,
(in an open-response box) the factors that influence their deci- explicit recommendation, loyalty program membership,
sions about whether to rate a review as helpful or unhelpful. On reviewer expertise, number of images uploaded with the
the next survey page, we showed them the description that they review, explicit mention of another brand in the review, and
had written and asked them whether they wrote that they were date of review (b ¼ .056, SE ¼ .014, p < .001; see Table 2).2
more likely to rate an Amazon review as helpful when the In summary, Study 5 provides evidence that Sephora users
review made them want to follow the reviewer’s advice (a rate reviews that mention a purchase mistake as more helpful.
measure of persuasion; Berry, Butler, and De Rosis 2005;
Burns 1991; Magee and Kalyanaraman 2010). A chi-square
2
For descriptive statistics of the variables included in Table 2, see Web
Appendix G. For a regression analysis on raw number of helpfulness votes,
1
A supplementary analysis on only this reduced set of reviews from the see Web Appendix H. For a discussion and regression analysis of an extended
manual coders yielded the same pattern of results as those presented in the model that predicts a Sephora-specific outcome (i.e., the number of “loves”
main text. Web Appendix F presents this regression analysis. included on the product page on the Sephora website), see Web Appendix I.
Reich and Maglio 61

Table 2. Results of Regression Analysis on Helpfulness Index for not been marred by mistakes. We integrate the persuasion and
Study 5. attribution theory literature streams to illuminate a powerful
Predictor b SE t p
factor that shapes the persuasive impact of consumer reviews
on purchase decisions.
(Constant) .083 .024 3.456 .001**
Mistake reference .056 .014 .171 4.136 .000***
Number of words .001 .000 .144 4.291 .000*** Alternative Explanations
Valence of review .018 .017 .044 1.027 .305
Star rating .005 .017 .023 .305 .761
In addition to illuminating the persuasive impact of mistake
Explicit recommendation .029 .028 .077 1.052 .293 makers and the mechanism underlying this phenomenon, we
Loyalty program member .014 .013 .035 1.109 .268 also examined several alternative explanations. Merely com-
Reviewer expertise .191 .230 .026 .828 .408 paring two brands proved insufficient to evoke our effect in
Number of uploaded images .023 .035 .021 .648 .517 Study 3, which instead provided evidence consistent with H4,
Mention of another brand .011 .020 .017 .552 .581 which posited domain dependence. The posttest of Study 3
Date of review 2.672E-5 .000 .009 .236 .813 provided further evidence inconsistent with the possibility
Notes: The R-square of the simple model (without controls) is .053; the that inferences regarding mistaken reviewers being discern-
R-square of the full model with controls is .080. ing, having integrity, being likeable, and being similar to the
*p < .05. message recipient contribute to the phenomenon we docu-
**p < .01.
***p < .001. ment, nor did potential surprise from admitting a mistake.
We note that these findings do not preclude the possibility
that some types of purchase mistakes could alter perceptions
Even after including numerous controls, the relationship of reviewers (for these or other traits) in a manner that would
between mention of a mistake and review helpfulness still affect their persuasive power. Nevertheless, the fact that the
holds. We targeted helpfulness as a meaningful construct of proposed phenomenon emerged—for both purchase decisions
consideration, as our pilot test and previous research indicate in the main study and inference of expertise in the posttest of
that a helpful review is a persuasive review, and marketing Study 3—when the reviewers were perceived equivalently on
managers interested in increasing sales begin with the goal of these dimensions suggests that these accounts are insufficient
persuading consumers to purchase their products. Thus, these to explain our findings.
results underscore the applied relevance of mentioning mis- Furthermore, the mere delivery of negative information
takes by providing evidence for a robust connection to review does not appear to increase mistaken reviewers’ persuasive
helpfulness using data taken from a real online retailer. impact (neither through loss aversion [Kahneman and Tversky
1979] nor through alternate routes including but not limited to
a negativity bias, a blemishing effect, mistake-induced per-
General Discussion ceptions of warmth, or negativity-induced perceptions of
People are often skeptical about whether the reviews they competence [Ein-Gar, Shiv, and Tormala 2012; Sen and
encounter are authored by well-informed consumers and thus Lerman 2007]). If the mere delivery of negative information
first evaluate whether a reviewer is credible before deciding drove mistaken reviewers’ persuasive impact, then the refer-
whether to rely on his or her review (e.g., Price and Stone 2004; ence to a mistake in a different product domain could
Wen, Tan, and Chang 2009; Woodside and Davenport 1974). have increased persuasion. We did not observe this outcome.
We find that people are more likely to conclude that a reviewer These results are thus inconsistent with the possibility that
has more expertise, and are thus more likely to purchase the the mere delivery of negative information increases mistaken
product that a reviewer recommends, if the reviewer admits to reviewers’ persuasive impact.
having made a purchase mistake in that domain. The current These findings also cannot be accounted for by a pratfall
research thus suggests that featuring purchase mistakes in effect (clumsy actions that enhance the attractiveness of super-
online consumer reviews offers a promising opportunity as a ior others; Aronson, Willerman, and Floyd 1966). For one
persuasive tactic. Our research therefore provides important, thing, this alternative possibility is theoretically implausible:
practical insight into the inputs underlying consumers’ deci- pratfalls only affect assessments of superior, and thus poten-
sions about whether to purchase reviewed products while also tially threatening, others (Aronson et al. 1966), and it seems
making several theoretical contributions. First, whereas sub- unlikely that participants in the current studies perceived their
stantial research has documented the negative inferences that fellow consumers to have a superiority needed for the pratfall
observers make about mistake makers, our research uncovers effect to emerge. Moreover, Study 3 finds that the mere pres-
the conditions in which learning about others’ mistakes leads ence of a mistake is insufficient to increase a reviewer’s per-
people to perceive those others more positively (cf. products suasive impact, as would be predicted by the pratfall effect;
made by mistake; Reich, Kupor, and Smith 2017)—in partic- thus, the current results are inconsistent with this potential
ular, the conditions in which mistaken reviewers are perceived alternative explanation. Taken together, these results thus sug-
to be more expert and better able to identify the optimal course gest that acknowledging mistakes can play an important role in
of action than even reviewers whose previous experience has promoting persuasion and influencing purchase decisions.
62 Journal of Marketing 84(1)

Future Directions moderate observers’ attributions. In the current studies, the


consequences of the admitted mistakes were relatively
The inferential process documented in these studies suggests
minor—for example, in Study 1, the reviewer’s mistake
that there are likely several boundaries to the persuasive power
merely resulted in the purchase of suboptimal headphones.
of mistakes. We documented one in the current research (H4):
People may perceive that mistakes that cause more severe
when a reviewer makes a mistake in a different domain from
consequences signal stable incompetence: because people
the focal purchase (Study 3), potential purchasers are no longer
assume that more severe outcomes emanate from larger ante-
more persuaded by the reviewer. Our theoretical framework
cedents (LeBoeuf and Norton 2011), they may further assume
predicts several other boundaries deserving of further attention
that larger antecedents are more foreseeable and attribute
as well. First, we have considered what happens when a prior foreseeable mistakes to incompetence. We encourage future
purchase mistake is corrected—presumably through the pro- research to examine these possibilities.
cess of gaining expertise—in a subsequent purchase. However, If the incompetent occupy the low end of the expertise spec-
if an initial mistake goes uncorrected (as in the case of a trum and reviewers who admit to their previous mistakes find
reviewer who mentions making not one but two successive themselves situated much higher along that same spectrum,
mistakes), then a lack in ostensibly gained expertise should what might determine where other reviewers fall on the basis
result in a lack of reviewer persuasiveness. This might suggest, of their reviews? Our research, in complementing existing
for instance, a boundary around review valence, as a negative work (Packard and Berger 2017), attests to the continued
product review would reasonably be more likely to include importance of addressing this question, as inferred expertise
mention of a (second) mistake than would a positive review. increases the likelihood that consumers will heed the advice
Second, mistaken reviewers may indeed be seen to have gained of reviews (as either featured by brands or encountered on
expertise, but they may no longer be more persuasive if people review repositories). For future research consideration, we
discount the utility of that expertise. In particular, not all mis- underscore one noteworthy domain in which consumers
takes lead to learning that is relevant for other consumers. For regularly have the opportunity to learn about products: that
example, suppose a person buys a speaker system and subse- of purchase decisions made by friends and family. In all of our
quently realizes that the purchase was a mistake because the experiments, reviewers acknowledge their own mistakes on the
color of the speakers does not match the person’s home decor. basis of direct prior experience: they made the purchase, used
That person may well have learned something from that expe- the product firsthand, became aware of its shortcomings, and
rience, but the expertise gained from this mistake bears no gained sufficient expertise to identify this purchase as a mis-
relevance to others whose homes are decorated with different take and to rectify it in a subsequent purchase.
color schemes. If so, admitting to making this kind of subjec- Incorporating the vantage point of mistaken purchases made
tive, idiosyncratic mistake may not increase a reviewer’s per- by others, we identify situations in which consumers might also
suasive impact because others may recognize the irrelevance of gain new information. First, a consumer might proceed through
the mistake (and thus the irrelevance of the mistaken all of the steps in the aforementioned sequence, save for mak-
reviewer’s subsequent expertise) to their own decisions. This ing the purchase in the first place. That is, the consumer might
would suggest more broadly that our effect should hold more borrow a product purchased by a friend and then proceed to use
strongly for more global, general reviews of products (but see and hold a negative evaluation of it. Would this consumer learn
Study 1). anything? We propose that it depends on the preexisting opin-
Our theorizing further suggests that if consumers attribute a ion held by (or at least acknowledged by) the consumer. If the
reviewer’s purchase mistake to stable incompetence that can- consumer knew all along that their friend’s purchase was a dud
not be fixed, they will be less likely to infer that expertise has and using it only verified this opinion, then summarizing this
been gained and may thus be less influenced by the mistaken event in a review should not signal any learning on the part of
reviewer’s product reviews. Accordingly, future research could the consumer authoring the review. Such a review might be
profit from investigating the factors that affect people’s exper- appraised in a manner similar to the reviewers in our studies
tise attributions. For example, observers may judge that an who made two successful purchases without any mention of a
admitted mistake is attributable to stable incompetence if the mistake—as stable, rather than trending upward in expertise—
observers perceive that the correct product choice was patently and would observe a degree of persuasiveness akin to theirs.
obvious at the time of the mistake. In such cases, observers may However, the consumer instead might have held high or neutral
infer that the cause of the mistake is most likely to be an expectations for the product borrowed from the friend, only to
absence of basic common sense rather than a fixable lack of have them dashed after using the product firsthand. Should this
knowledge about the product category. Extending the common type of consumer write a review recapping their experience, we
sense thread, should the content of the review contain not only expect that it would convey that learning had taken place (i.e., a
mention of a mistake but also other content indicative of a lack knowledge-based update to the reviewing consumer’s degree
of expertise (e.g., stating that a newly purchased speaker serves of expertise possible even in the absence of making the mis-
as a fantastic paperweight, using poor grammar), our model taken purchase per se), which our results suggest is key in
would predict that the reviewer’s advice would have less translating the admission of a mistake (be it a mistaken pur-
impact. The severity of a mistaken outcome may further chase or simply a mistaken opinion) into powerful persuasion.
Reich and Maglio 63

Perhaps reviewers who use products purchased by others As previously noted, companies frequently feature con-
and describe how they learned from the experience would be sumer reviews to market their products. Thus, in addition to
highly persuasive but still not as persuasive as reviewers who providing novel insight into the inferences that people make
made the purchase mistake themselves. This possibility, echo- about others’ purchase mistakes, this research also has signif-
ing the aforementioned possibility of a continuum or spectrum icant practical import because it suggests that marketing man-
of reviewer persuasiveness, might arise as the result of our agers may strategically omit information that actually increases
proposal that the admission of a personal purchase mistake is persuasive influence. In other words, featuring reviews that
more costly (and more diagnostic of confident expertise) than include purchase mistakes might be a widely underused per-
the admission of a personal incorrect (favorable) opinion about suasive tactic: in their attempts to increase their persuasive
a product. To be sure, should the friend in this situation (rather influence, managers may inadvertently decrease persuasive
than the consumer doing the borrowing) write a review attest- influence through the missed opportunity of failing to feature
ing to their own purchase mistake, our results suggest this mistakes. The present investigation thus offers the clear direc-
review would be maximally persuasive. But what if, instead, tive to incorporate more (perhaps any) mention of mistakes
the friend merely described their experience to the consumer when featuring reviews to promote products.
and the consumer then authored a review summarizing the We propose that this directive may be especially well-suited
friend’s experience—would the reviewing consumer (rather to smaller firms with fewer marketing-related resources at their
than the friend) have an impact on potential purchasers? We disposal. What the effect documented by the present investiga-
offer that the answer to this question might depend on the tion lacks in magnitude and everyday prevalence, it makes up
degree to which those potential purchasers see the reviewing for in subtlety and ease of implementation. As such, it may help
consumer as socially close to the friend who made the mis- level the playing field between large corporations that can pour
take. If the relationship seems distant, then readers might significant resources into market research, carefully determin-
discount the potential for the reviewer to learn from the ing which reviews to feature, and smaller companies that lack
friend’s mistake. But, if the relationship seems close, then such deep market research pockets. These relatively smaller
readers might believe that the reviewer gained just as much companies may take comfort in (and win sales from) the insight
expertise as the mistaken friend (Kang, Hirsh, and Chasteen that featuring a review referencing a prior mistake acts as a
2010), bolstering the degree to which potential purchasers simple but beneficial tool in shaping consumer preference.
place stock in the review. Firms are not alone in their desire to persuade, and our
findings not only might warrant consideration by the marketing
departments of large corporations but also might be brought to
Contributions to Practice bear on any attempt to convince others to take purchase-related
The primary takeaway of our research for practitioners advises advice. Dovetailing with the literature on word of mouth (Babić
the featuring of mistakes to drive more online traffic and, ulti- Rosario et al. 2016; Berger 2014; Maglio and Reich 2019;
mately, more sales. As such, outlets at which online retailers Packard and Berger 2017), we offer two additional domains
have control to structure the decision environment provide the well-suited to apply our work. First, friends often give
best point of entry for this recommendation. Though, to be sure, purchase-related advice to each other, driven either by the
they cannot control the content of online reviews authored by relatively selfless joy of facilitating a positive purchase expe-
independent consumers, online retailers do have the power to rience for their friend (e.g., a great gym) or by more self-
flag certain reviews as “featured” or “highlighted,” warranting interested motives (e.g., a gym that promises a referral bonus).
placement ahead of an otherwise long and undifferentiated list We propose that both objectives should be facilitated by mak-
of reviews. By identifying one or multiple reviews that mention ing mention of a previous mistake, extending our work to closer
a previous purchase mistake and bumping them up to the top, interpersonal relationships.
online retailers can make online shoppers more likely to see, Second, a growing number of individuals have taken to
read, and accept the advice of these reviews that our research online forums to build personal brands around product reviews
suggests prove especially persuasive. However, it is not only and tutorials, as evidenced by the thousands of personal blogs
online retailers that aspire to put helpful reviews in front of reviewing electronics and YouTube channels demonstrating
their audience. Review curation websites benefit from persuad- how to apply makeup. Insofar as these influencers hope to build
ing customers not toward any one particular course of their personal brands in the form of likes, follows, and men-
purchase-related action but, instead, toward feeling that the tions, they need to be seen as credible experts. Our model offers
website itself offers a valuable source of information. If cus- the possibility that perhaps they would be more likely to attain
tomers believe that sites such as Yelp, TripAdvisor, and Rotten their objectives should they incorporate mention of previous
Tomatoes offer a consistent, reliable set of reviews, they return mistakes into their content. Aside from the content of the
to them more frequently ahead of various purchases, with such review, they might also persuade others to click on their written
increased traffic in turn increasing advertising revenues for and recorded reviews in the first place by including mention of
curation sites. Broadly speaking, then, any firm or brand in the a mistake in the title itself (e.g., “Learn from my mistake!”).
business of offering helpful, positive advice should feature Only after navigating to their review will others read and incor-
reviews that mention previous purchase mistakes. porate the reviewer’s purchase-related advice. We note,
64 Journal of Marketing 84(1)

though, that the decision to follow an influencer on social Berry, Dianne C., Laurie T. Butler, and Fiorella De Rosis (2005),
media in perpetuity results from a confluence of many factors “Evaluating a Realistic Agent in an Advice-Giving Task,” Inter-
that may or may not overlap with the momentary evaluation of national Journal of Human-Computer Studies, 63 (3), 304–27.
helpfulness and one-time purchase decisions around which the Burns, Robert B. (1991), “Persuasion by Communication,” in Essen-
present investigation centered. Still, this application of the phe- tial Psychology: For Students and Professionals in the Health and
nomenon documented and detailed herein would suggest that Social Services. Lancaster, UK: Kluwer Academic Publishers,
inferred expertise, through admission of mistakes, can not only 236–54.
drive sales but also build brand equity writ large. Campion, Michael A. and Robert G. Lord (1982), “A Control Systems
Conceptualization of the Goal-Setting and Changing Process,”
Associate Editor Organizational Behavior and Human Performance, 30 (2),
Donna Hoffman 265–87.
Carver, Charles S. and Michael F. Scheier (2012), Attention and
Self-Regulation: A Control-Theory Approach to Human Behavior.
Declaration of Conflicting Interests Berlin: Springer Science & Business Media.
The author(s) declared no potential conflicts of interest with respect to Chesney, Thomas and Daniel Su (2010), “The Impact of Anonymity
the research, authorship, and/or publication of this article. on Weblog Credibility,” International Journal of Human-
Computer Studies, 68 (10), 710–18.
Funding Chevalier, Judith A. and Dina Mayzlin (2006), “The Effect of Word of
The author(s) received no financial support for the research, author- Mouth on Sales: Online Book Reviews,” Journal of Marketing
ship, and/or publication of this article. Research, 43 (3), 345–54.
Dellarocas, Chrysanthos (2003), “The Digitization of Word of Mouth:
Promise and Challenges of Online Feedback Mechanisms,” Man-
References
agement Science, 49 (10), 1401–24.
Alba, Joseph W. and J. Wesley Hutchinson (1987), “Dimensions of Dweck, Carol S. and Ellen Leggett (1988), “A Social-Cognitive
Consumer Expertise,” Journal of Consumer Research, 13 (4),
Approach to Motivation and Personality,” Psychological Review,
411–54.
95 (2), 256–73.
Angelis, Matteo De, Andrea Bonezzi, Alessandro M. Peluso, Derek D.
East, Robert, Kathy Hammond, and Malcolm Wright (2007), “The
Rucker, and Michele Costabile (2012), “On Braggarts and Gossips:
Relative Incidence of Positive and Negative Word of Mouth: A
A Self- Enhancement Account of Word-of-mouth Generation and
Multi-Category Study,” International Journal of Research in
Transmission,” Journal of Marketing Research, 49 (4), 551–63.
Marketing, 24 (2), 175–84.
Arcillo, Maria (2011), “The Elements of a Testimonial Approach
Ein-Gar, Danit, Baba Shiv, and Zakary Tormala (2012), “When
Type of Television Advertisement That Makes It Appealing to
Blemishing Leads to Blossoming: The Positive Effect of Nega-
Consumers,” research report, University of Santo Tomas (accessed
tive Information,” Journal of Consumer Research, 38 (5),
July 18, 2017), http://docslide.net/documents/the-elements-of-a-
846–59.
testimonial-approach-type-of-television-advertisement-that-
Gawande, Atule (1999), “Annals of Medicine: When Doctors Make
makes-it-appealing-to-consumers.html.
Mistakes,” The New Yorker (February 1), https://www.newyorker.
Argyris, Chris and Donald Schon (1978), Organizational Learning: A
Theory of Action Approach. Reading, MA: Addison-Wesley. com/magazine/1999/02/01/when-doctors-make-mistakes.
Aronson, Elliot, Ben Willerman, and Joanne Floyd (1966), “The Grewal, Dhruv, Sukumar Kavanoor, Edward F. Fern, Carolyn Costley,
Effect of a Pratfall on Increasing Interpersonal Attractiveness,” and James Barnes (1997), “Comparative Versus Noncomparative
Psychonomic Science, 4 (6), 227–28. Advertising: A Meta-Analysis,” Journal of Marketing, 61 (4),
Babić Rosario, Ana, Francesca Sotgiu, Kristine De Valck, and Tammo 1–15.
H.A. Bijmolt (2016), “The Effect of Electronic Word of Mouth on Harper, Elizabeth (2014), “Here Are the Best Ways You Can Lever-
Sales: A Meta-Analytic Review of Platform, Product, and Metric age Positive Online Reviews,” Sprout Social (March 10), https://
Factors,” Journal of Marketing Research, 53 (3), 297–318. sproutsocial.com/insights/how-to-leverage-positive-reviews-into-
Baek, Hyunmi, JoongHo Ahn, and Youngseok Choi (2012), social-success/.
“Helpfulness of Online Consumer Reviews: Readers’ Objectives Ho-Dac, Nga N., Stephen J. Carson, and William L. Moore (2013),
and Review Cues,” International Journal of Electronic Commerce, “The Effects of Positive and Negative Online Customer Reviews:
17 (2), 99–126. Do Brand Strength and Category Maturity Matter?” Journal of
Baker, Sara M. and Richard E. Petty (1994), “Majority and Minority Marketing, 77 (6), 37–53.
Influence: Source-Position Imbalance as a Determinant of Mes- Hong, Ying-yi, Chi-yue Chiu, and Carol S. Dweck (1995), “Implicit
sage Scrutiny,” Journal of Personality and Social Psychology, Theories of Intelligence,” in Efficacy, Agency, and Self-Esteem,
67 (1), 5–19. Michael H. Kernis, ed. New York: Springer, 197–216.
Berger, Jonah (2014), “Word of Mouth and Interpersonal Communi- Huei-Chen, Hsu (2007), “A Study of Factors Affecting the Success of
cation: A Review and Directions for Future Research,” Journal of Private Label Brands in Chinese e-Market,” Journal of Technology
Consumer Psychology, 24 (4), 586–607. Management in China, 2 (1), 38–53.
Reich and Maglio 65

Kahneman, Daniel and Amos Tversky (1979), “Prospect Theory: An Peterson, Christopher, Amy Semmel, Carl Von Baeyer, Lyn Y.
Analysis of Decision Under Risk,” Econometrica, 47 (2), 263–91. Abramson, Gerald I. Metalsky, and Martin E.P. Seligman (1982),
Kang, Sonia K., Jacob B. Hirsh, and Alison L. Chasteen (2010), “Your “The Attributional Style Questionnaire,” Cognitive Theory and
Mistakes Are Mine: Self-Other Overlap Predicts Neural Response Research, 6 (3), 287–99.
to Observed Errors,” Journal of Experimental Social Psychology, Podsakoff, Philip M. and Jiing-Lih Farh (1989), “Effects of Feedback
46 (1), 229–32. Sign and Credibility on Goal Setting and Task Performance,”
Kats, Rimma (2018), “When Buying Expensive Items, Consumers Organizational Behavior and Human Decision Processes, 44 (1),
Turn to Reviews,” eMarketer (January 26), https://www.emark 45–67.
eter.com/content/when-buying-expensive-items-consumers-turn- Price, Paul C. and Eric R. Stone (2004), “Intuitive Evaluation of Like-
to-reviews. lihood Judgment Producers: Evidence for a Confidence Heuristic,”
Kruger, Justin, Derrick Wirtz, Leaf Van Boven, and T. William Journal of Behavioral Decision Making, 17 (1), 39–57.
Altermatt (2004), “The Effort Heuristic,” Journal of Experimental Rawlins, Anthony (2011), “Why Including Consumer Reviews on
Social Psychology, 40 (1), 91–98. Travel Sites Matters,” Phocus Wire (June 28), https://www.phocus
Kunda, Ziva (1999), Social Cognition: Making Sense of People. Cam- wire.com/Why-including-consumer-reviews-on-travel-sites-
bridge, MA: MIT Press. matters.
Kupor, Daniella, Taly Reich, and Kristin Laurin (2018), “The Reich, Taly, Daniella M. Kupor, and Rosanna K. Smith (2017), “Made
(Bounded) Benefits of Correction: The Unanticipated Interperso- by Mistake: When Mistakes Increase Product Preference,” Journal
nal Advantages of Making and Correcting Mistakes,” Organiza- of Consumer Research, 44 (5), 1085–1103.
tional Behavior and Human Decision Processes, 149 (11), Sen, Shahana and Dawn Lerman (2007), “Why Are You Telling Me
165–78. This? An Examination into Negative Consumer Reviews on the
Laband, David N. (1986), “Advertising as Information: An Empirical Web,” Journal of Interactive Marketing, 21 (4), 76–94.
Note,” Review of Economics and Statistics, 68 (3), 517–21. Spencer, Steven J., Mark P. Zanna, and Geoffrey T. Fong (2005),
LeBoeuf, Robin and Norton, Michael I. (2011), “Consequence-Cause “Establishing a Causal Chain: Why Experiments Are Often More
Matching: Looking to the Consequences of Events to Infer their Effective Than Mediational Analyses in Examining Psychological
Causes,” Journal of Consumer Research, 39 (1), 128–41. Processes,” Journal of Personality and Social Psychology, 89 (6),
Lipsman, Andrew, Rebecca Chadwick, Cindy Liu, Karin von Abrams, 845–51.
and Yoram Wurmser (2019), “US 2018 Holiday Season Review Sun, Baohong, Scott A. Neslin, and Kannan Srinivasan (2003),
and 2019 Preview,” eMarketer (February), https://www.emarketer. “Measuring the Impact of Promotions on Brand Switching When
com/content/us-2018-holiday-season-review-and-2019-preview. Consumers Are Forward Looking,” Journal of Marketing
Locke, Edwin A., Karyll N. Shaw, Lise M. Saari, and Gary P. Latham Research, 40 (4), 389–405.
(1981), “Goal Setting and Task Performance: 1969–1980,” Van Trijp, Hans C.M., Wayne D. Hoyer, and J. Jeffrey Inman (1996),
Psychological Bulletin, 90 (1), 125–52. “Why Switch? Product Category: Level Explanations for True
Magee, Robert G. and Sriram Kalyanaraman (2010), “The Perceived Variety-Seeking Behavior,” Journal of Marketing Research, 33
Moral Qualities of Web Sites: Implications for Persuasion Pro- (3), 281–92.
cesses in Human–Computer Interaction,” Ethics and Information Walter, Clyde K. and John R. Grabner (1975), “Stockout Cost Models:
Technology, 12 (2), 109–25. Empirical Tests in a Retail Situation,” Journal of Marketing, 39
Maglio, Sam J. and Taly Reich (2019), “Feeling Certain: Gut Choice, (3), 56–60.
the True Self, and Attitude Certainty,” Emotion, 19 (5), 876–88. Wen, Chen, Bernard C.Y. Tan, and Klarissa Ting-Ting Chang (2009),
Mudambi, Susan M. and David Schuff (2010), “What Makes a Helpful “Advertising Effectiveness on Social Network Sites: An Investi-
Review? A Study of Customer Reviews on Amazon.com,” MIS gation of Tie Strength, Endorser Expertise and Product Type on
Quarterly, 34 (1), 185–200. Consumer Purchase Intention,” in Proceedings of the International
Murphy, Patrick E. and Ben M. Enis (1986), “Classifying Products Conference on Information Systems. Atlanta: Association for
Strategically,” Journal of Marketing, 50 (3), 24–42. Information Systems, 151–71.
Murphy, Rosie (2017), “Local Consumer Review Survey 2017,” Wojnicki, Andrea C. and David Godes (2017), “Signaling Success:
BrightLocal (accessed October 6, 2019), https://www.brightlocal. Word of Mouth as Self-Enhancement,” Customer Needs and Solu-
com/learn/local-consumer-review-survey-2017/. tions, 4 (4), 68–82.
Packard, Grant and Jonah Berger (2017), “How Language Shapes Woodside, Arch G. and J. William Davenport (1974), “The Effect of
Word of Mouth’s Impact,” Journal of Marketing Research, 54 Salesman Similarity and Expertise on Consumer Purchasing
(4), 572–88. Behavior,” Journal of Marketing Research, 11 (2), 198–202.
Parker, Dianne and Rebecca Lawton (2003), “Psychological Contri- Yeomans, Michael (2019), “Some Hedonic Consequences of
bution to the Understanding of Adverse Events in Health Care,” Perspective-Taking in Recommending,” Journal of Consumer Psy-
Quality and Safety in Health Care, 12 (6), 453–57. chology, 29 (1), 22–38.

You might also like