Abstract
An emerging virtual-reality technology, virtual fitting rooms (VFRs) allow online shoppers to virtually try on clothes. Despite its increasing popularity, how VFR technology influences different consumer groups is hitherto unknown. Neglecting such nuances may significantly undermine VFR effectiveness. From a large-scale field experiment with real-world transactional data and five laboratory experiments, the authors document the asymmetric effects of VFR use conditional on consumer body types, characterize the theoretical underpinnings, and identify a systematic set of managerially actionable moderators that can mitigate adverse effects. Specifically, while VFR use enhances product evaluations and purchases among consumers with relatively low body mass index (BMI) levels, it negatively influences responses from high-BMI consumers due to self-image threat induced by avatars that resemble consumers’ own bodies. To cope with self-image threat, high-BMI consumers tend to shift the blame to the apparel item, resulting in negative product responses. The authors identify four feasible solutions to alleviate the negative responses among high-BMI users of VFRs, namely, promoting diversified beauty standards, featuring mannequin faces for VFR avatars, providing consumers opportunities to engage in prosocial behavior, and presenting high-status products. These findings offer guidance for retailers to leverage this new technology smartly to enhance both business performance and consumer well-being.
The apparel industry is an important economic sector that contributes almost 2% of global gross domestic product (Statista 2022). Sales from online fashion retail are projected to top U.S. $1 trillion by 2025, with a 7.18% compounded annual growth rate (Orendorff 2022). To compete for a share of this growing market, e-tailers are exploring new ways to make online shopping more interactive. A technological tool that has garnered significant interest recently is the virtual fitting room (VFR) or virtual showroom, an interactive simulation technology that allows customers to virtually try clothes on avatars or virtual models, using texture-mapped product images. Narrowing the gap between online and offline shopping experiences, VFRs have helped the apparel industry recoup customers during the COVID-19 pandemic, as many people hesitate to return to brick-and-mortar stores and physically try on clothes (Lee 2020). For these reasons, an increasing number of fashion brands and retailers have adopted VFR technology, including Adidas, Hugo Boss, Macy's, Ted Baker, and Under Armour (Biron 2020).
Prior research has largely found that VFRs enhance apparel sales (Yang and Xiong 2019); however, whether this positive effect is universal across consumer groups is unknown. As a personalized VFR exposes users to virtual avatars resembling their own body shape, we propose that its effect is likely dependent on user body mass index (BMI), a key measure of body size commonly used to classify individuals (Algars et al. 2009; Stice and Whitenton 2002). By highlighting users’ body size, VFRs may induce a threat to the self-image of overweight consumers, consequently lowering their product evaluations and decreasing purchases.
Although overweight consumers constitute the majority (nearly 74%) of the U.S. adult population (Fryar, Carroll, and Afful 2020), they have traditionally been overlooked by fashion retailers. Industry reports indicate that overweight customers are often treated “like an anomaly” in retail stores, and they are more inclined to shop online to avoid unpleasant experiences (Anthony 2017). While VFR technology seems to be a useful tool for online shopping, our research cautions that it has the potential to damage overweight consumers’ feelings of self-worth. The psychological well-being of overweight consumers is currently a critical issue, with accumulated evidence connecting overweight with depression and even suicidality (Haynes et al. 2019). We show that instead of helping high-BMI consumers, VFRs may trigger and exacerbate the harmful psychological impact of being overweight, which consequentially leads to less favorable product evaluations and fewer purchases. Importantly, this research goes beyond identifying the detrimental repercussions of VFR use and proposes actional managerial strategies that alleviate self-image threat and avert high-BMI consumers’ negative responses to apparel products.
From a large-scale field experiment (Study 1) with real-world transactional data, we find that the effect of VFR use on product purchases becomes negative as user BMI increases. A follow-up laboratory experiment (Study 2) reveals self-image threat to be the underlying mechanism for this phenomenon. Studies 3–6 identify four theoretically and managerially relevant boundary conditions under which the adverse effect of VFR on high-BMI consumers is alleviated or reversed: (1) when consumers are primed with the diversified beauty concept (vs. idealized beauty standard), (2) when VFR avatars feature a mannequin face (vs. consumers’ own faces), (3) when consumers are provided (vs. not provided) an opportunity to engage in prosocial behaviors, and (4) when the apparel product is perceived to be of high (vs. low) status. Figure 1 depicts the overarching conceptual framework.
Conceptual Framework.
This research contributes to the VFR literature in two main regards. First, we document an important consumer heterogeneity effect of VFR use: while the technology produces a positive effect among low-BMI consumers, it tends to induce self-image threat among high-BMI consumers that reduces their product evaluations and purchases. Second, our research suggests ways to safeguard high-BMI consumers’ self-image and identifies a systematic set of instruments to attenuate the adverse effects of VFR use on product outcomes.
Literature on Virtual Fitting Rooms
Despite their obvious importance in apparel retail, empirical research on fitting rooms, physical or virtual, is scarce (Yang and Xiong 2019; Yun, Jung, and Choo 2015). Employing small-scale experiments or interviews, existing studies on physical fitting rooms have largely focused on how design elements or functionalities influence fitting-room evaluations and overall shopping experiences (e.g., Baumstarck and Park 2010; Seo and Fiore 2016). Regarding VFR technology (see summary of literature in Table 1), previous studies have examined VFRs with a standardized virtual model featuring the same body figure for all users (Fiore, Kim, and Lee 2005; Lee, Fiore, and Kim 2006). In keeping with technological advancement, recent studies have moved toward exploring personalized VFRs that use a customized virtual model based on users’ own body figures and faces. For example, Kim and Forsythe (2007, 2008) show that the perceived usefulness, entertainment, and ease of use of the technology influence users’ attitudes to personalized VFRs and to the retail website.
Summary of Relevant Literature.
Our research follows the most recent literature and examines the impact of personalized VFRs (except for Study 4, which also examines “semipersonalized” VFRs, in which the avatar's body shape is personalized but its face is not). Two previous studies are particularly relevant to our research. First, Merle, Senecal, and St-Onge (2012) found that body esteem positively influences consumers’ perceived self-congruity with the virtual model (i.e., whether the model resembles their own body image), which further enhances perceptions of the technology and shopping intentions. This finding hints at the theoretical possibility that consumers’ body size could influence VFR effectiveness.
Second, as an exception to most existing studies that have focused on user attitudes toward VFRs or to an online store, Yang and Xiong (2019) document that VFR use increases product sales via enhanced shopping enjoyment and reduced purchase risk when consumers are exposed to the technology as the sole product display. They also found that a personalized VFR generates more sales than a nonpersonalized VFR. However, the positive effect of a personalized (vs. nonpersonalized) VFR is reversed when users are simultaneously presented a photo of the product worn by a fashion model, which prompts social comparison and induces self-discrepancy.
While prior studies imply the potential roles of body image and self-feelings, they have not investigated whether or how the impact of VFRs on product evaluations and purchases may differ across consumer segments. Because VFRs highlight the user's own body image, it is important to examine their effects contingent on consumers’ body types. While the VFR literature is mute regarding any effect of consumer heterogeneity, this research fills the gap by revealing diverging VFR effects across consumers with varying BMI levels. Furthermore, we contribute to the literature by identifying a systematic set of strategies to mitigate adverse VFR effects among high-BMI consumers. Instead of merely alerting retailers to the potential dark side of VFRs, our findings suggest actionable solutions to optimize the benefits of VFRs when targeting various consumer segments.
Main Proposition and Its Underlying Mechanism
VFRs and Self-Image Threat
Human beings have a fundamental need to possess and maintain a positive view of the self (Steele 1988). Adverse experiences that present negative feedback to one's self-image threaten such needs, often resulting in heightened psychological discomfort or distress (Steele, Spencer, and Lynch 1993). Marketing researchers have documented that self-image threat can be induced by a diverse array of events, including experiencing social exclusion (Mourey, Olson, and Yoon 2017), being subject to negative stereotypes (Amatulli et al. 2018; Lee, Kim, and Vohs 2011), and perceiving loss of control (Chen, Lee, and Yap 2017). Following prior work, we define self-image threat as “an experience that calls into question one's favorable views about oneself” (Steinhart and Jiang 2019, p. 742).
In our context, VFRs may present a threat to high-BMI consumers’ self-image by providing unfavorable feedback about their physical attractiveness. Today's society largely endorses an idealistic standard of physical appearance that closely relates beauty with slenderness (e.g., Richins 1991). Accordingly, the stigma associated with being overweight is prevalent (Brownell et al. 2005), and a larger body size is often regarded as less attractive, a notion held by both overweight and nonoverweight individuals (Major, Eliezer, and Rieck 2012). While the societal emphasis on beauty tends to exert more pressure on women, recent research demonstrates that the ideal of thinness influences both male and female consumers’ evaluations of their own physical attractiveness (Argo, Dahl, and Adaval 2018).
VFRs construct digital avatars according to users’ own body measurements, and for overweight users, such virtual models may serve as a reminder of their body imperfections and the standard of slimness that they fail to meet. While heightened saliency of body-image deficiencies is threatening to self-perception in general (Fredrickson et al. 1998), it is especially so when individuals’ body shapes are captured visually, for example in photos and videos. Kiefer, Sekaquaptewa, and Barczyk (2006) find that participants experienced heightened feelings of threat and performed worse in a test with peers when they viewed photos of themselves in which they appeared wider (manipulated via photo retouching). Major, Eliezer, and Rieck (2012) show that overweight participants became more concerned about their body image and made more effort to cope with the psychological distress induced by the threat when their speech was videotaped (vs. audiotaped) and used for evaluations regarding the first impression they would make. In a similar vein, we propose that viewing a VFR avatar that resembles one's own body would pose a threat to the self-image of high-BMI consumers, who tend to view their bodies as less attractive. Due to such self-image threat, these individuals would in turn generate distinct responses to apparel products, compared with the responses of consumers with lower BMI.
BMI and VFR Effects on Product Outcomes
When receiving information that poses a threat to self-image, individuals tend to experience psychological discomfort and are motivated to eliminate such negative feelings (Dunning 2007; Mandel et al. 2017). With a need to maintain a positive self-view, consumers generally resist the threatening feedback and, instead, often resort to responses that rationalize the information to defend their self-image (Sherman and Cohen 2006). By interpreting the unfavorable event as reflecting deficiencies in the external environment rather than internal attributes of the self-concept, the threat is ameliorated and the self-image is protected (Dunning and Cohen 1992). The use of motivated rationalization has been illustrated in marketing contexts where, to preserve their self-image, consumers encountering threat shift the blame from themselves to the product. For example, Dunn and Dahl (2012) find that consumers who experience product failure due to their own inadequacies, when given an opportunity to complain, tend to rate the product lower. Dahl, Argo, and Morales (2012) show that consumers denigrate an apparel product when feeling threatened by witnessing another more attractive consumer wearing the same product.
Drawing on this theory, we propose that high-BMI users of VFRs are motivated to engage in defensive processing to cope with threat and protect their self-image. Rather than attributing the imperfect VFR image to themselves, they tend to blame the apparel product instead. Such attribution, while enhancing the self-image of high-BMI VFR users, likely leads to their unfavorable responses toward the product. This detrimental effect, however, may be meager among low-BMI users, who are less likely to experience significant threat. Instead, VFR use may generate positive product evaluations and purchases among the latter group because of the benefits it brings to their shopping process, such as enhanced enjoyment and reduced risk (Yang and Xiong 2019). Taking these factors together, we hypothesize:
Boundary Conditions
In this section, we discuss the boundary conditions whereby the adverse impact of VFRs on high-BMI consumers’ product responses may be mitigated. We focus on four moderators that are of both theoretical and managerial significance. If self-image threat indeed drives unfavorable responses of high-BMI consumers as we theorize in H2, reducing the perception or impact of the threat should be essential to alleviating the damaging VFR consequences. Hence, we identify moderators that are effective in curbing self-image threat, based on the literature (Mandel et al. 2017; Sherman and Cohen 2006). Specifically, we show that the adverse effect of VFRs among high-BMI consumers tends to be mitigated by (1) resolving the source of the threat (by highlighting diversified beauty standards), (2) dissociating the consumer from the trigger of the threat (by featuring a mannequin face on VFR avatars), (3) affirming consumers’ self-image before the threat (by engaging them in prosocial behavior), and (4) repairing self-image after the threat (by purchasing high-status products). By establishing their mitigation of the adverse VFR effect among high-BMI consumers, we provide further support for our theory and for the central role of self-image threat. Moreover, our proposed moderators are all actionable/manipulatable for marketers and thus offer meaningful managerial implications.
Beauty Norm: Resolving the Sources of the Threat
Self-image threat tends to be reduced in situations where consumers can directly address the source of the threat (“direct resolution”; Mandel et al. 2017). Accordingly, we propose that a diversified beauty norm will mitigate self-image threat and moderate the negative effect of VFRs.
As theorized, the self-image threat induced by VFRs among high-BMI consumers is underlined by the beauty ideal that often equates thinness with physical attractiveness (Richins 1991). As long as high-BMI consumers hold themselves to such a standard of slimness, they are likely to view their own physique as unattractive and thus feel threatened by VFR avatars that mirror their own body shape. While the idealized beauty standard remains prominent today, growing social effort is being made to construct a definition of beauty that is more diversified and inclusive (Pounders 2018). Some fashion brands have joined this movement and started promoting the diversity of beauty in ad campaigns that feature models of various body sizes, ages, and/or skin colors. These efforts promote the belief that everyone is beautiful in their own way and that no one-size-fits-all standard exists. Such a diversified beauty concept, once primed, can reduce consumers’ internalization of the thinness ideal (Bissell and Rask 2010). When the pressure to conform to an idealized beauty standard is reduced, high-BMI consumers should be less likely to feel threatened when viewing their personalized avatars and thus less motivated to shift blame to the apparel product. Hence, we anticipate the negative effect of VFRs on high-BMI consumers’ product responses to dissipate when diversified beauty is primed.
VFR Avatars with Mannequin Face: Dissociation from the Threat
As another effective coping strategy, “dissociation” separates consumers from the negative event to alleviate self-image threat (Mandel et al. 2017). Hence, we propose that using a mannequin face for VFR avatars can reduce self-image threat and the adverse VFR effect.
To create a vivid and lifelike try-on experience, most current VFR applications feature personalized avatars based on customers’ own faces and bodies (Yang and Xiong 2019), and we use this type of VFR avatars in most of our tests. While personalized body figures allow users to visually assess apparel fit, featuring users’ own facial images strengthens their perception that the avatar is a representation of their own appearance. For high-BMI users who view the avatar as providing negative feedback, we argue that VFRs featuring their own faces can be counterproductive. Prior research has documented that facial image is a strong identity cue (Rhodes and Leopold 2011) and viewing one's own face elevates self-consciousness (Maister and Tsakiris 2013). As a VFR avatar with a user's face highlights self-identity, thereby increasing consumer association with the avatar, high-BMI customers are more likely to view such virtual models as an embodiment of their own body imperfections. By contrast, displaying a generic mannequin-like face creates some distance between a consumer and the VFR avatar. As consumers become less likely to identify with the avatar or take its image personally, VFRs may no longer pose as severe a threat to their self-image. Hence, we anticipate that compared with avatars based on users’ own faces, those with a mannequin face are less likely to induce a negative impact of VFRs among high-BMI consumers.
Prosocial Behavior: Self-Affirmation Before Threat
The self-affirmation literature has accumulated much evidence that bolstering global self-worth can reduce the threat associated with negative feedback and shield individuals from potential damages (Sherman and Cohen 2006; Steele 1988). When personal values in domains unrelated to the threat are reinforced, individuals’ self-evaluations become more stable and less vulnerable to negative feedback. Accordingly, we anticipate that affirming consumers’ self-worth before exposure to VFR images would effectively mitigate self-image threat and the negative VFR effect.
Specifically, we propose that providing consumers with opportunities to engage in prosocial acts would affirm important personality traits such as kindness and generosity (Bizman et al. 1980) that help buttress their self-image to fend off the threat induced by VFR use. Prosocial acts are voluntary behaviors intended to benefit others, and are universally regarded as desirable (Aknin, Van de Vondervoort, and Hamlin 2018). Prior research has shown that merely thinking about engaging in prosocial behavior is sufficient to increase consumers’ positive self-concepts (Khan and Dhar 2006). With a strengthened sense of worth, high-BMI consumers’ self-image is less likely to be bruised by their physical appearance as spotlighted by VFRs. Therefore, allowing consumers to engage in prosocial behavior can reduce the likelihood for high-BMI consumers to feel threatened and to denigrate a product under consideration, and the negative influence of VFRs should thus dissipate.
Product Status: Repairing Self-Image After Threat
Next, we examine a product characteristic that allows consumers to cope with the VFR threat and restore damaged self-image. According to Mandel et al. (2017), individuals may cope with self-image threat by engaging in behavior that enhances a different aspect of their self-image distinct from the threat (e.g., wealth and social status, “fluid compensation”). Hence, we expect high product status to mitigate or even reverse VFRs' negative effect.
Specifically, we propose that although in general high-BMI consumers tend to denigrate the apparel item after experiencing a threat to protect their self-image, products with high status could be a notable exception to the negative VFR effect. This proposition is consistent with findings in the literature that status purchase signals buyers’ wealth and social class, and is thus often sought out to compensate for the threat and repair self-image (Song, Huang, and Li 2017). In situations where consumers receive negative self-image feedback (e.g., low cognitive capabilities, powerlessness), previous research shows an increase in both purchase intentions and willingness to pay for high-status products (Rucker and Galinsky 2009; Sivanathan and Pettit 2010). Under the self-image threat posed by VFRs, high-BMI consumers tend to generate unfavorable responses to low-status products as a result of engaging in motivated rationalization to protect a positive self-view. Conversely, however, self-image threat will likely make high-status products appear more attractive due to their self-enhancing capability (Mandel et al. 2017). Motivated to take actions to restore their self-image damaged by VFRs, high-BMI consumers will likely find high-status products more appealing and generate more positive product responses. Hence, we expect that the negative VFR effect among high-BMI consumers would persist for low-status products but be reversed for high-status products.
Overview of Studies
We conduct six studies to test our hypotheses. Study 1 is a large-scale field experiment conducted in an online apparel store from which we obtained real purchase data. This study tests H1 and investigates the differential effects of VFRs on apparel sales across customers of varying BMI levels. Study 2 replicates the findings of Study 1 in a laboratory setting and examines the mediating role of self-image threat, testing H2. Studies 3–6 test H3 through H6 using lab experiments, which examine the moderating roles of beauty norm (Study 3), mannequin face (Study 4), prosocial behavior (Study 5), and product status (Study 6) and provide additional evidence for the self-image threat account. Table 2 provides an overview of the six studies and our main findings.
Summary Statistics and Key Findings of Each Study.
Notes: Entries are means (standard deviations in parentheses) unless otherwise specified. DDD = difference-in-difference-in-differences; DV = dependent variable; ME = mediator. We also conducted Studies A1 and A2 (Web Appendix C) to examine the moderating roles of need for uniqueness and self-affirmation and find that the negative impacts of VFRs on product responses of high-BMI customers through self-image threat are mitigated by salient need for uniqueness or self-affirmation.
Study 1: Field Experiment
To assess the effects of VFRs on product sales across consumer segments, we conducted a field experiment in a large online women's apparel store in China. 1 On August 25, 2016, the retailer launched VFRs for a randomly selected subset of its products (307 of 917 products; randomization checks presented subsequently). After inputting body measurements (e.g., height, weight, bust, waist, hips) and uploading a full-face photo, a customer can virtually try on products in VFRs with an avatar based on their own face and body shape (see example in Figure 2). The launch of VFRs created an opportunity for us to test the treatment effect of VFRs on sales by comparing the products for which VFR was made available (the treatment group) with the remaining products (the control group).
Example of VFR Try-On Image in Study 1.
We obtained customer purchase data from the retailer over eight weeks from July 28 to September 21, that is, four weeks before the launch of the VFR (pretreatment period) and four weeks after (posttreatment period). During the eight weeks, there was no other major change in the online store. To calculate customers’ BMI levels, we obtained customers’ height and weight information from the retailer's customer database. 2 Following the definition of the World Health Organization (1995), an individual's BMI is computed as weight in kilograms divided by the square of height in meters. After removing 877 customers whose body size information was unavailable or ambiguous (i.e., having multiple inconsistent records of weight or height in the database), the final sample consists of 29,985 product purchases made by 8,233 customers.
We performed randomization checks to ensure that product items were randomly assigned to the treatment group (with VFRs) and the control group (without VFRs). As shown in Table WA2 in Web Appendix A, the observed percentage of treated products within each category (top, bottom, and dress) falls within the 95% confidence interval (CI) of the corresponding category, indicating that the allocation of the treated and untreated products is proportionately applied to all product categories. Table WA3 in Web Appendix A shows no significant differences between the treatment group and the control group in terms of product price and average sales volume before the experiment. Hence, differences in treatment across the two groups are less likely to be attributable to idiosyncratic attributes of the products.
Estimating the Treatment Effect
To examine the treatment effect of VFRs on product sales and its moderators, we specify a difference-in-differences (DID) model following the literature (Janakiraman, Lim, and Rishika 2018):
Results
Column 1 of Table 3 presents the estimation results of Equation 1 with robust standard errors clustered at the product level. 5
Results of Study 1 (Field Experiment).
**p < .01. ***p < .001.
Notes: Results in Column 1 are estimated from Equation 1 with product-level data (as products were randomly assigned to treatment and control groups), where the main effect of Treat is omitted because product-specific fixed effects are controlled for. Results in Column 2 are estimated from Equation WA3 in Web Appendix A as a robustness check with customer-level data and BMI as a continuous variable, where the main effects of Treat and BMI and their interaction term Treat × BMI are omitted because customer-specific fixed effects are controlled for. Robust standard errors, shown in parentheses, are clustered at the product level in Column 1 and at the consumer level in Column 2.
Main effect of VFRs
We find a positive and significant treatment effect of VFRs on product sales (coefficient of Treat × Post α2 = 3.49, p < .001). The positive main effect of VFRs is consistent with that reported by Yang and Xiong (2019).
Moderating effect of BMI
H1 proposes that the effect of VFRs varies with customer BMI. The coefficient of Treat × Post × BMI (α6 = −.44, p < .001) suggests that BMI negatively moderates the treatment effect of VFRs. Figure 3 visualizes this moderating effect of BMI: the treatment effect of VFRs decreases as user BMI increases and turns negative for high-BMI customers (i.e., BMI groups 8–10, whose BMI ≥ 23). Hence, we find support for H1. Results remain robust using customer-level data (Column 2 of Table 3; see Web Appendix A for model estimation details). Web Appendix A also provides additional robustness checks and falsification tests.
Treatment Effect of VFR Conditional on BMI.
In summary, based on field experiment data, Study 1 shows that VFRs do not universally increase product purchases; instead, the effect varies across customer groups. Specifically, the effect of VFRs on product purchases is moderated by customer BMI: it is positive for low-BMI customers only, but negative among high-BMI customers, thus supporting H1.
Study 2: Underlying Mechanisms
Participants, Design, and Procedure
Study 2 tests H1 in a controlled lab experiment to increase internal validity. It also explores self-image threat as the underlying mechanism and tests H2. The study employed a 2 (product display: VFR vs. product photo) × BMI between-subjects design, in which product display was manipulated and BMI was calculated from self-reported height and weight. Participants were recruited from a large university in the United States in exchange for course credits. Prior to the experiment, participants completed a preliminary questionnaire, in which they reported their body measurements (height and weight), indicated body shape by selecting one of five images (pear, inverted triangle, apple, rectangle, or hourglass) that best matched their own (Makhanya et al. 2014), and uploaded a high-resolution front-facing photo. We used the information and photos to create personalized VFR avatars for the experiment. We also collected data on control variables including gender, age, online shopping frequency, and familiarity with VFRs. A total of 130 participants (65.38% female, 34.62% male; Mage = 20.17 years, SD = 1.52) provided valid responses and were emailed a web link to the experiment several weeks later.
In the experiment, participants imagined shopping online for two-piece outfits (a dress shirt and pants) for an upcoming job interview. In the VFR condition, participants viewed an image of their own personalized avatar wearing the outfits. In the control condition, participants viewed a simple photograph of the dress shirt and pants (see Web Appendix B for product image stimuli in lab studies).
Subsequently, participants evaluated the outfits on a four-item scale (1 = “negative,” and 7 = “positive”; 1 = “unfavorable,” and 7 = “favorable”; 1 = “undesirable,” and 7 = “desirable”; 1 = “disliked,” and 7 = “liked”; Argo, Dahl, and Adaval 2018, α = .96). To measure self-image threat, we asked participants to report how viewing the picture made them feel on a four-item scale adapted from Mende et al. (2019) (e.g., “The picture makes people like me less important,” “The picture threatens my worthiness”; 1 = “strongly disagree,” and 7 = “strongly agree”; α = .92). Since prior research has documented perceived shopping enjoyment (α = .97) and risk (r = .88) as drivers of the positive VFR effect (Yang and Xiong 2019), we measured both variables to replicate previous findings. We also collected additional control variables following the literature, including expected product performance (“How do you expect the product to look on you?”; 1 = “unattractive,” and 7 = “flattering”; Eckman, Damhorst, and Kadolph 1990) and self-avatar similarity (“To what extent does the model in the try-on picture accurately represent you?” “To what extent does the model in the try-on picture accurately represent your body shape?”; 1 = “very inaccurately,” and 7 = “very accurately”; Merle, Senecal, and St-Onge 2012). Web Appendix B lists the key constructs and their measurements in all lab studies.
Results
Hypothesis testing
We ran regressions with VFRs (yes vs. no), BMI (continuous), their interaction, and control variables to predict product evaluations. 6 The results revealed a significant main effect of VFRs (β = 2.93, t(121) = 2.61, p = .01). Consistent with H1, the interaction effect between VFRs and BMI was negative and significant (β = −.13, t(121) = −2.73, p = .007). We plot the interaction effect and compute Johnson–Neyman points using floodlight analysis (Spiller et al. 2013). As shown in Figure 4, Panel A, there is a positive simple effect of VFRs on product evaluations for low-BMI participants (BMI < 24.63) and a negative effect of VFRs for high-BMI participants (BMI > 37.71), supporting H1.
Plots for Study 2.
Mediation analyses
To examine the underlying mechanism, we ran PROCESS Model 7 (Hayes 2018) with 5,000 bootstrap samples, with self-image threat, perceived shopping enjoyment, and risk as mediators and all control variables included (we also ran PROCESS Model 8 as a robustness check in Studies 2–4 and found consistent results). Figure 4, Panel B, visualizes moderated mediation analysis results. The overall moderated mediation effect of our focal mediator, self-image threat, is marginally significant (β = −.02, SE = .02, 90% CI: [−.06, −.001]), as the indirect effect through self-image threat is significant for high-BMI participants only (β = −.14, SE = .10, 95% CI: [−.39, −.001]), but not for low-BMI ones 7 (β = −.01, SE = .06, 95% CI: [−.12, .15]). For purposes of parsimony, results for the other mediators (enjoyment and risk) are reported in Web Appendix B, where we also present ancillary mediation analysis to rule out additional alternative mediators (e.g., avatar similarity).
In sum, Study 2 not only replicates the field experiment findings of Study 1 that VFRs produce divergent effects among consumers with varying BMI levels, but also demonstrates self-image threat as the mediator driving the negative impact of VFRs on high-BMI consumers. In the following studies, we focus on this adverse VFR effect and examine moderators that could effectively mitigate it.
Study 3: Idealized Versus Diversified Beauty Standards
Participants, Design, and Procedure
Study 3 tests H3 regarding the moderating role of beauty norms in a controlled lab setting. The study employed a 2 (beauty norm: diversified vs. idealized) × 2 (body size of VFR avatar: true vs. enlarged) + 1 (control) between-subjects design. To increase internal validity, instead of measuring participants’ BMI levels, we manipulated their body size perceptions in this study by altering VFR avatars to reflect participants’ true versus enlarged body size. In addition to the four experimental conditions, we included a control condition in which participants received no beauty-norm manipulation and viewed a plain product photo instead of a VFR image, to establish a baseline of product responses for comparison purposes. Two hundred fifty-seven students (78.60% female, 21.40% male; Mage = 20.63 years, SD = 2.03) recruited from a large university in China completed a preliminary survey to report body measurements and information on control variables (i.e., gender, age, online shopping frequency, and familiarity with VFRs).
Several weeks after the preliminary survey, participants took part in the experiment in a lab on an individual basis in exchange for 15 Chinese yuan (about two U.S. dollars) and were randomly assigned to the five conditions. To obtain participants’ photos to create personalized virtual models, we informed them upon their arrival at the lab that they would evaluate a digital camera on the basis of pictures of themselves taken with the camera. We then took a full-body, front-facing photo of each participant, which was used to create the personalized VFR avatar in the backstage. Subsequently, participants completed a questionnaire with instruments for the focal study.
In the questionnaire, all participants imagined shopping online for a new pair of jeans on a retail website and viewed an online ad, which included beauty-norm manipulation (see images in Web Appendix B). In the diversified beauty-norm condition, the ad featured ten female models varying in body sizes and ages and a tagline: “Be beautiful in your own way.” In the idealized beauty-norm condition, the ad showed ten young and slim female models and a tagline: “Be beautiful. Be envy of the party.” After viewing the ad, participants were asked to write down an example that was consistent with the ad message.
Participants in the experimental conditions then imagined using a VFR to virtually try on the product and viewed a picture of their personalized avatar wearing the jeans (in the control condition, participants viewed the product photo without a VFR avatar). We manipulated body size perception by varying the avatar's body size, following Kiefer, Sekaquaptewa, and Barczyk (2006). In the true body-size condition, the avatar was based on the original photo of the participant. In the enlarged size condition, the photo was retouched to be 20% wider to create the avatar. To check whether the body-size manipulation worked as intended, we conducted a posttest among participants in the experimental conditions, who reported perceptions of their body sizes in the VFR avatars (i.e., “How do you feel your body looks in the image?”; 1 = “skinny,” and 7 = “plump”). A 2 (beauty norm) × 2 (body size) analysis of variance (ANOVA) revealed only a main effect of the body-size manipulation. Supporting the manipulation, participants in the enlarged (vs. true) size condition perceived their body size in the VFR image as larger (Menlarged = 4.92, SD = 1.36; Mtrue = 3.53, SD = 1.53; F(1, 200) = 46.87, p < .001, partial η2 = .19).
Afterward, participants evaluated the jeans on the same scale as in Study 2 (α = .95) and rated their purchase intention on a three-item scale (e.g., “I feel like buying the jeans now,” “I would like to buy the jeans as soon as possible,”; 1 = “strongly disagree,” and 7 = “strongly agree”; Yang and Xiong 2019; α = .98). We then measured the level of self-image threat (α = .83) and perceived shopping enjoyment (α = .81) using the same scales as in Study 2. Subsequently, all participants in the experimental conditions rated perceived similarity of the VFR avatar on a scale similar to that in Study 2 (r = .57). To check the beauty-norm manipulation, we asked participants to report the extent to which they adopted the idealized beauty norm on a three-item scale (e.g., “Thin women are more attractive than curvy women;” 1 = “strongly disagree,” and 7 = “strongly agree”; Ramati-Ziber, Shnabel, and Glick 2020; α = .84). Finally, we collected data on additional control variables, including perceived novelty, involvement, and interactivity of the product display. Before leaving the lab, participants were debriefed that they no longer needed to evaluate the digital camera. We recorded the time length that each participant took to complete the questionnaire, which was included as another control variable in data analyses.
Results
Manipulation check
A 2 (beauty norm) × 2 (body size) ANOVA on participants’ adoption of the idealized beauty norm revealed only a main effect of the beauty-norm manipulation, and participants in the diversified (vs. idealized) beauty-norm condition conformed to the idealized beauty standard to a lesser degree (Mdiversified = 3.94, SD = 1.08; Midealized = 4.29, SD = 1.10; F(1, 200) = 5.22, p = .02, partial η2 = .03), supporting the manipulation.
Hypothesis testing
In the four experimental conditions, we estimated 2 (beauty norm) × 2 (body size) analysis of covariance models on the two dependent variables, product evaluation and purchase intention, respectively, with all control variables included. Consistent with H3, both models revealed a significant two-way interaction effect (on product evaluation: F(1, 192) = 4.48, p = .04, partial η2 = .02; on purchase intention: F(1, 192) = 5.36, p = .02, partial η2 = .03).
When the idealized beauty norm was primed, participants in the enlarged body-size condition had significantly less favorable product responses (product evaluations: Menlarged = 2.66, SD = 1.15; purchase intention: Menlarged = 1.76, SD = .86) than those in the true body-size condition (product evaluations: Mtrue = 3.73, SD = 1.40; F(1, 192) = 14.07, p < .001, partial η2 = .07; purchase intention: Mtrue = 2.74, SD = 1.57; F(1, 192) = 13.42, p < .001, partial η2 = .07) and those in the control condition (product evaluations: Mcontrol = 3.86, SD = 1.27; F(1, 95) = 18.73, p < .001, partial η2 = .17; purchase intention: Mcontrol = 2.56, SD = 1.23; F(1, 95) = 11.42, p = .001, partial η2 = .11). In addition, product responses under the true body-size condition and the control condition are comparable (product evaluations: F(1, 94) = .11, p = .74, partial η2 = .001; purchase intention: F(1, 94) = 1.06, p = .31, partial η2 = .01), suggesting that the general positive VFR effect dampened when the idealized beauty norm was salient.
However, when the diversified beauty norm was primed, the difference between the two body-size conditions dissipated (product evaluation: Mtrue = 3.92, SD = 1.25; Menlarged = 3.57, SD = 1.30; (1, 192) = .37, p = .54, partial η2 = .002; purchase intention: Mtrue = 2.81, SD = 1.24; Menlarged = 2.69, SD = 1.28; F(1, 192) = .05, p = .82, partial η2 = .000). Further, for participants who viewed an enlarged-size VFR avatar, their responses were significantly enhanced when the diversified versus idealized beauty norm was primed (product evaluations: Mdiversified = 3.57, SD = 1.30; Midealized = 2.66, SD = 1.15; F(1, 192) = 15.55, p < .001, partial η2 = .07; purchase intention: Mdiversified = 2.69, SD = 1.28; Midealized = 1.76, SD = .86; F(1, 192) = 16.39, p < .001, partial η2 = .08). These findings confirmed the mitigation effect of priming a diversified beauty norm.
Mediation analyses
We ran PROCESS Model 7 with 5,000 bootstrap samples with self-image threat and shopping enjoyment as mediators and all control variables included. Results showed a significant moderated mediation effect of self-image threat (product evaluations: β = .38, SE = .15, 95% CI: [.10, .70]; purchase intention: β = .20, SE = .11, 95% CI: [.03, .45]). Importantly, we found indirect effects of diversified (vs. idealized) beauty norm through self-image threat for participants in the enlarged body-size condition (product evaluations: β = .31, SE = .12, 95% CI: [.10, .56]; purchase intention: β = .17, SE = .08, 95% CI: [.04, .36]). However, for participants in the true-size condition, the indirect effects through self-image threat dissipated (product evaluations: β = −.07, SE = .09, 95% CI: [−.24, .10]; purchase intention: β = −.04, SE = .05, 95% CI: [−.15, .05]), supporting H3. Web Appendix B visualizes the mediation analysis and reports the results pertaining to shopping enjoyment along with ancillary analyses to rule out avatar similarity and time length as alternative mediators.
In sum, Study 3 shows that when an idealized beauty norm is salient, enlarged-size avatars in VFRs lead to less favorable product responses, driven by self-image threat. This effect disappears when a diversified beauty norm is primed, consistent with H3. Adopting the belief that all body shapes are beautiful effectively removes self-image threat, leading consumers exposed to enlarged-size VFR avatars to generate product responses as favorable as those exposed to nonenlarged avatars.
Study 4: VFR Avatar with User's Face Versus Mannequin Face
Participants, Design, and Procedure
Study 4 tests H4 regarding the moderating role of the avatar's face in VFRs, using a 3 (product display: VFR avatar with user's own face vs. VFR avatar with mannequin face vs. product photo) × BMI between-subjects design. Product display was manipulated, and BMI was calculated from height and weight. Students recruited from a large university in China completed a preliminary survey and reported measures of control variables (gender, age, online shopping frequency, and familiarity with VFRs) and body measurements (height, weight, and body shape). One hundred fifty-one students (79.47% female, 20.53% male; Mage = 19.64 years, SD = 1.80) participated in the lab study on an individual basis in exchange for course credits.
Similar to Study 3, we informed participants upon their arrival at the lab that they would evaluate a digital camera, and then we took full-body, front-facing photos of them. Afterward, participants completed a questionnaire containing instruments for the focal study. They imagined shopping online for a two-piece outfit (a dress shirt and pants) for an upcoming interview. In the two VFR conditions, participants read that they used a VFR to try on the outfits and viewed an image of their personal avatars wearing the products. While the body figure of the VFR avatar was based on the body measurements and the full-body photo of each participant, the avatar showed either the participant's own face (in the user's face condition) or a standard mannequin face (in the mannequin face condition). In the control condition, participants viewed an image of the outfits only. To control for the time used to process the visual information, all participants were shown the image for ten seconds.
Afterward, participants evaluated the outfits (α = .91) and reported perceived self-image threat (α = .88) on the same scales as in Studies 2 and 3. We then measured mood state as a control variable (“How positive or negative is your mood right now?”; 1 = “very negative,” and 7 = “very positive”). As manipulation checks, participants in the VFR conditions rated the avatar's face similarity, body-shape similarity, and overall similarity. Before leaving the lab, participants were debriefed that they no longer needed to evaluate the camera.
Results
Manipulation checks
A one-way ANOVA on the three similarity scores showed that participants in the user's face (vs. mannequin face) avatar condition perceived a higher level of face similarity (Muser = 4.69, SD = 2.03; Mmannequin = 2.84, SD = 1.76; F(1, 101) = 24.44, p < .001, partial η2 = .19) and a higher level of overall similarity (Muser = 3.56, SD = 1.94; Mmannequin = 2.27, SD = 1.59; F(1, 101) = 13.42, p < .001, partial η2 = .12), but the two groups did not differ on body shape similarity (Muser = 3.81, SD = 1.66; Mmannequin = 3.33, SD = 1.62; F(1, 101) = 2.16, p = .15, partial η2 = .02). Thus, the manipulation was successful.
Hypothesis testing
We created two dummy variables for the three experimental conditions: mannequin face (mannequin face = 1, user's face = 0, control = 0) and control group (control = 1, user's face = 0, mannequin face = 0). We then conducted a regression analysis on product evaluations with the dummy variables, BMI, and the interactions between BMI and each dummy variable as independent variables. All control variables were included as covariates. Results revealed a main effect of BMI (β = −.24, t(140) = −4.41, p < .001), a main effect of the mannequin face dummy (β = −6.06, t(140) = −3.06, p = .003), a main effect of the control group dummy (β = −4.74, t(140) = −2.36, p = .02), an interaction effect between BMI and the mannequin face dummy (β = .29, t(140) = 3.21, p = .002), and an interaction effect between BMI and the control group dummy (β = .22, t(140) = 2.32, p = .02).
Floodlight analysis on the BMI × mannequin face interaction (Web Appendix B) shows that a VFR avatar with the user's face (vs. mannequin face) led to more favorable product evaluations for low-BMI users (BMI < 18.55), but less favorable evaluations for high-BMI participants (BMI > 22.41). This finding supports H4 that, for high-BMI consumers, VFR effectiveness increases when featuring avatars with a mannequin face instead of users’ own faces. Floodlight analysis on the BMI × control interaction shows that compared with the control condition, a VFR avatar with the user's face increased product evaluations for low-BMI participants (BMI < 19.75) but decreased those for high-BMI participants (BMI > 25.95). This finding replicates the results in Studies 1–3.
Mediation analyses
To test the mediating role of self-image threat, we ran PROCESS Model 7 with 5,000 bootstrap samples (Hayes and Preacher 2014) in which the experimental condition was the independent variable (user's face as baseline) with control variables included. We found a significant moderated mediation effect of VFR avatar with mannequin (vs. user's) face through self-image threat (β = .09, SE = .05, 95% CI: [.01, .21]). There was a marginally significant indirect effect of mannequin (vs. user's) face through self-image threat for high-BMI participants (β = .26, SE = .16, 90% CI: [.01, .54]), but not for low-BMI participants (β = −.20, SE = .18, 90% CI: [−.52, .06]).
The results also supported a significant moderated mediation effect of the control condition (vs. VFR with user's face) through self-image threat (β = .09, SE = .04, 95% CI: [.02, .18]). As expected, the indirect effect of the control condition (vs. VFR with user's face) through self-image threat was significant for high-BMI participants (β = .26, SE = .14, 95% CI: [.02, .57]), but not for low-BMI ones (β = −.17, SE = .15, 95% CI: [−.49, .11]). Ancillary analyses ruled out mood, face similarity, body similarity, and overall similarity as viable mediators (see Web Appendix B).
In sum, Study 4 shows that feelings of self-image threat are effectively lowered among high-BMI consumers when VFR avatars feature a mannequin face (vs. users’ own faces), resulting in enhanced product evaluations. Hence, we find support for H4.
Study 5: Prosocial Behavior
Participants, Design, and Procedure
In Study 5, we test H5 regarding the moderating role of prosocial behavior. The study used a 2 (product display: VFR vs. product photo) × 2 (prosocial behavior: yes vs. no) × BMI between-subjects design, in which we manipulated product display and prosocial behavior and calculated participants’ BMI levels. Since the focal product in this study was a dress, we recruited only female students from a university in China. Participants first completed a preliminary survey and reported body measurements (height, weight, and body shape) and measures of control variables (age, online shopping frequency, and familiarity with VFRs). Several weeks later, 203 female students (Mage = 21.26 years, SD = 1.95) took part in the experiment in a lab on an individual basis in exchange for course credits.
We used a similar cover story (evaluation of a digital camera) as in Studies 3 and 4 and took a full-body, front-facing photo of each participant. Participants then completed the focal study, where they imagined shopping for a dress on a retail website. In the prosocial-behavior condition, participants read the website's promotional offer that 10% of their purchase would be donated to support education for children in poverty. In the control condition, the promotional offer consisted of giving back to customers 10% of their purchase in the form of a supermarket gift card. From prior research findings that thinking about hypothetical virtue behavior increases one's positive self-view (Kiefer, Sekaquaptewa, and Barczyk 2006), we expected that participants who read about donating 10% of the purchase (vs. receiving a gift card with 10% of the purchase value) were more likely to experience affirmation of self-worth.
Subsequently, participants viewed either an image of their personal avatar wearing the dress (the VFR condition) or a display of only the dress (the product photo condition) and evaluated the dress on an item adapted from Argo, Dahl, and Adaval (2018) (1 = “negative,” and 7 = “positive”). Using the same scales as in previous studies, we measured self-image threat (α = .90) and control variables, including mood state, perceived novelty, involvement, and interactivity of the product display. To check the prosocial-behavior manipulation, participants rated whether their purchase from the website would help others (1 = “strongly disagree,” and 7 = “strongly agree”). Before leaving the lab, participants were informed that they no longer needed to evaluate the digital camera.
Results
Manipulation check
A 2 (product display) × 2 (prosocial behavior) ANOVA on participants’ perceptions of the purchase revealed only a main effect of the prosocial-behavior manipulation, and participants in the prosocial-behavior condition (vs. no prosocial behavior) showed stronger agreement that purchasing from the website would help others (Mprosocial = 5.32, SD = 1.15; Mno prosocial = 3.21, SD = 1.44; F (1, 199) = 131.20, p < .001, partial η2 = .40). Hence, the manipulation was successful.
Hypothesis testing
We ran a regression on product evaluations with VFR (yes vs. no), prosocial behavior (yes vs. no), BMI, and their two-way and three-way interactions as independent variables, with control variables included. The results showed a main effect of VFR (β = 9.91, t(188) = 4.28, p < .001), an interaction effect between VFR and BMI (β = −.45, t(188) = −4.23, p < .001), and a three-way interaction effect of VFR, BMI, and prosocial behavior (β = .36, t(188) = 2.18, p = .03).
To facilitate the interpretation of the three-way interaction, we ran separate regression analyses for the prosocial and control conditions, with VFR, BMI, and their interaction term as independent variables, including all control variables. We found a significant VFR × BMI interaction effect in the no-prosocial-behavior condition (β = −.50, t(92) = −3.80, p < .001). Floodlight analysis (Web Appendix B) shows that, compared with displaying a product photo, VFR enhanced product evaluations for low-BMI participants (BMI < 20.77) but reduced evaluations for high-BMI ones (BMI > 24.00), replicating previous findings. However, in the prosocial-behavior condition, there was only a significant and positive effect of VFR (β = 5.01, t(89) = 2.39, p = .02), while the VFR × BMI interaction dissipated (β = −.12, t(89) = −1.23, p = .22). This result is in line with H5 that VFR is more likely to generate positive product outcomes among high-BMI consumers when they are given opportunities to engage in prosocial behavior.
Mediation analysis
To test the mediating role of self-image threat, we ran PROCESS Model 11 with 5,000 bootstrap samples (Hayes 2018), including all the covariates. The overall effect of the model was significant (β = .09, SE = .06, 95% CI: [.001, .22]). In the no-prosocial-behavior conditions, the moderated mediation effect through self-image threat was significant (β = −.07, SE = .04, 95% CI: [−.17, −.004]). Specifically, the indirect effect of VFR (vs. product photo) through self-image threat was significant among high-BMI participants (β = −.18, SE = .12, 95% CI: [−.48, −.001]) but not among low-BMI ones (β = .15, SE = .14, 95% CI: [−.06, .50]), replicating previous findings. In the prosocial-behavior conditions, the moderated mediation effect via self-image threat was no longer significant (β = .02, SE = .03, 95% CI: [−.04, .09]) (the indirect effect of VFR through self-image threat was nonsignificant for either high- or low-BMI participants), further supporting H5.
In sum, consistent with H5, Study 5 shows that providing consumers with an opportunity to engage in prosocial behavior is effective in reducing the self-image threat that VFRs impose on high-BMI consumers, subsequently enhancing their product evaluations.
Study 6: Product Status
Participants, Design, and Procedure
Study 6 tests the moderating role of product status (H6). It employed a 2 (product display: VFR vs. product photo) × 2 (product status: high vs. low) × BMI between-subjects design, in which product display and status were manipulated and BMI levels were calculated. All students enrolled in a marketing course at a Chinese university were asked to complete a preliminary survey, in which they reported body measurements (e.g., height, weight, bust, belly, waist, hip, and bottom) and information on control variables (gender, age, and familiarity with VFRs). To ensure equivalent numbers of overweight and nonoverweight participants, we randomly invited 140 students with BMI ≥ 23 and another 140 students with BMI < 23 to participate in the study, following the literature that sets BMI of 23 as the borderline for being overweight among Asian consumers (Stewart, Dang, and Chen 2016). A total of 276 students (50.36% female, 49.64% male; Mage = 18.81 years, SD = .75) accepted the invitation and joined the study in the lab on an individual basis, for compensation of 30 Chinese yuan (about 4.5 U.S. dollars).
Using the same cover story as in the previous studies, we took a full-body front-facing photo of each participant upon their arrival. Participants then completed the focal study, in which they imagined shopping online and browsing a web page presenting a sweater. In the VFR condition, participants imagined that they used a VFR to try on the sweater and were shown an image of their personal avatar wearing the product. In the control condition, we showed the participants a photo of the sweater alone.
Product status was manipulated via product description displayed next to the visual image. In the high-status condition, the sweater was described as “handmade by designers,” “knitted from premium baby cashmere,” and “offering unparalleled luxury and timeless elegance.” In contrast, in the low-status condition, the sweater was depicted as “designed for everyday wear,” “knitted from a blend of cotton and nylon,” and “offering good value and a casual style.”
Afterward, participants evaluated the sweater on the same scale as in Studies 2–4 (α = .91). Because self-image threat often results in a temporary reduction of self-esteem according to prior research (e.g., Liu et al. 2017), we asked participants to report their perceived self-esteem levels at the present moment on Rosenberg's (1965) ten-item scale (e.g., “I am able to do things as well as most other people”; 1 = “strongly disagree,” and 7 = “strongly agree”; α = .95). We also collected data on additional control variables, including perceived novelty, involvement, and interactivity of the product display. Finally, to check the manipulation of product status, participants rated the perceived status of the sweater on three items (1 = “low status,” and 7 = “high status”; 1 = “common,” and 7 = “luxury”; 1 = “ordinary,” and 7 = “prestigious”; α = .93; Sivanathan and Pettit 2010). Before leaving the lab, participants were informed that they no longer needed to evaluate the digital camera.
Results
Manipulation check
A 2 (product display: VFR vs. product photo) × 2 (product status: high vs. low) ANOVA on perceived product status revealed only a significant main effect of the product status manipulation. The sweater was rated higher in perceived status in the high (vs. low) status condition (Mhigh status = 3.76, SD = 1.38; Mlow status = 2.83, SD = 1.00; F(1, 272) = 40.30, p < .001, partial η2 = .13). In addition, to check whether the product status manipulation inadvertently affected perceived product quality, we conducted a posttest in which 130 participants read either the high-status or low-status version of the product description before rating the perceived quality of the sweater on a five-item scale (e.g., “Compared with other sweaters, what is the likely quality of the sweater on the web page?”; 1 = “low quality,” and 7 = “high quality”; α = .89; Srivastava and Mitra 1998). Results showed no significant difference between the two conditions (Mhigh status = 5.06, SD = .72; Mlow status = 4.87, SD = .92; F(1, 128) = 1.72, p = .19, partial η2 = .01), thus ruling out perceived quality as a potential confound.
Hypothesis testing
We ran regression with VFR (yes vs. no), product status (high vs. low), BMI (continuous), and their two-way and three-way interactions as independent variables, including all control variables. Results showed a main effect of VFR (β = 4.10, t(262) = 2.38, p = .02), an interaction effect between VFR and BMI (β = −.19, t(262) = −2.59, p = .01), and, importantly, a three-way interaction effect between VFR, BMI, and product status (β = .32, t(262) = 3.16, p = .002).
Floodlight analyses (Web Appendix B) show that, when product status was low, VFR (vs. product photo) increased product evaluations for low-BMI participants (BMI < 20.75) but decreased product evaluations for high-BMI participants (BMI > 23.93), replicating previous results. When product status was high, VFR (vs. product photo) led to more positive evaluations among high-BMI participants (BMI > 22.14). These results support H6; that is, the negative VFR effect among high-BMI consumers is reversed when product status is high.
Mediation analysis
To examine the mediating role of self-esteem, we ran PROCESS Model 21 (Hayes 2018) with 5,000 bootstrap samples. All control variables were included. The results showed a significant overall effect (β = .12, SE = .05, 95% CI: [.04, .22]).
In the low-status condition, the moderated mediation effect was significant (β = −.06, SE = .03, 95% CI: [−.13, −.002]). Specifically, consistent with results in the other studies, a negative indirect effect through self-esteem emerged among high-BMI participants (β = −.34, SE = .17, 95% CI: [−.69, −.02]), but not among low-BMI participants (β = −.04, SE = .06, 95% CI: [−.19, .05]).
In the high-status condition, the moderated mediation effect was also significant (β = .06, SE = .03, 95% CI: [.02, .12]): we found a positive indirect effect through self-esteem among high-BMI participants (β = .34, SE = .12, 95% CI: [.12, .59]), but not among low-BMI participants (β = .04, SE = .05, 95% CI: [−.05, .15]). These findings provide further support for H6.
In sum, Study 6 showed that exposure to VFR avatars temporarily reduced high-BMI consumers’ self-esteem, which lowered their evaluations of a low-status product but enhanced evaluations of a high-status product. Thus, status products reverse the negative VFR effect among high-BMI consumers, as they provide a coping mechanism to manage and curtail the aftereffect of self-image threat. Considering that perceptions of product status are complex and often multiply determined (Kreuzbauer, King, and Basu 2013), one may argue that the specifics of the status manipulation used in this study (e.g., handmade) might have confounded our results. Thus, we replicated the study with additional analysis of the field data from Study 1, using a different operationalization of product status (Web Appendix A). Following the literature that price is an effective signal of product status (Lichtenstein, Ridgway, and Netemeyer 1993), we used price as a status proxy in the analysis. The results show that overweight customers’ reactions to VFRs were conditional on product price, and the adverse effect of VFRs on overweight customers’ purchases was reversed for high-price products. Thus, with different operationalizations of product status, we found converging support for H6.
General Discussion
This research provides a new conceptual framework and empirical evidence that advance the understanding of the effects of VFRs on product evaluations and purchases across different consumer groups. Two major findings emanate from the field and lab experiments. First, the launch of VFRs does not necessarily increase favorable product responses for all customers. VFR effectiveness is bounded by consumer body sizes and can be negative for high-BMI consumers because of self-image threat induced by virtual avatars representing their own body image. Second, the adverse effect of VFRs among high-BMI consumers can be mitigated by factors that reduce self-image threat or remedy its impact. When an individual is primed with diversified beauty norms or provided opportunities to engage in prosocial behavior, when VFR avatars feature a mannequin face, or when the apparel product is of high status, VFRs will no longer trigger negative product responses among high-BMI customers.
Contributions to the Literature
The VFR is a critical application in retail of the fast-developing virtual-reality technology. According to Goldman Sachs (2016), revenue from virtual-reality retail software alone may reach $1.6 billion in 2025. Despite its potential, VFR technology is understudied in the literature, and marketers are eager to understand how to better leverage it to enhance sales, especially as the COVID-19 pandemic has transformed the retail industry (Sheth 2020). Apparel retailers have faced unprecedented challenges, with many physical stores forced to close or offer contactless options. Highlighting the urgency of the matter and the timeliness of our research, an article in Vogue Business (McDowell 2020) stated, “With fitting rooms closed, physical retailers are looking to online tools to circumvent try-ons and increase consumer confidence.”
The limited literature on VFRs has largely focused on the benefits of the technology, and no prior study has examined contingency effects of VFRs across consumers or consumer segments (see Table 1). Neglecting such effects, marketers that adopt VFRs cannot achieve optimal performance and may even suffer significant losses. Our research provides a more complete understanding of VFR effects by identifying an important consumer-level moderator. From a fresh angle, we demonstrate how VFR technology acts as a double-edged sword that delivers mixed blessings to the online apparel industry: while it positively influences product evaluations and purchases among low-BMI customers, it can also induce unfavorable responses from high-BMI customers. This research goes beyond alerting e-tailers to this detrimental effect of VFRs to offer insights into how to mitigate or reverse it.
Our findings further indicate that VFRs can not only potentially harm e-tailers’ sales performance but also damage high-BMI consumers’ self-perceptions. Ruling out avatar similarity, mood, or information-processing time as alternative explanations, our results provide strong support for the role of self-image threat, highlighting VFRs' adverse effect on high-BMI users’ psychological well-being. We provide evidence that technological applications that spotlight consumer images may threaten self-evaluations, which can in turn affect consumers’ purchase responses to personal adornment items such as apparel.
According to the self-image threat account, we identify managerial actions that can be taken to mitigate adverse effects of VFRs. We document that the perception and impact of self-image threat can be decreased by removing the sources of the threat (beauty norm, Study 3), by dissociating the threat from the consumer (mannequin face, Study 4), or by affirming self-image in a different domain before the threat (prosocial behavior, Study 5). When the perceived threat is minimized, the negative VFR effects tend to dissipate. Furthermore, we show that, even after the threat has taken effect and damaged self-image, consumers actively seek opportunities to restore a positive self-perception, increasing their preferences for products with the capacity for self-image enhancement. Hence, for high-status products (Study 6), the negative effect of VFRs tends to reverse, resulting in more favorable product outcomes. Overall, these findings demonstrate the dynamic nature of how self-image threat affects consumer attitudes and behaviors, as well as the various measures that can be taken to manage and cope with such threatening feedbacks.
The literature has asserted that VFR effectiveness increases as the virtual model becomes more personalized. For example, Merle, Senecal, and St-Onge (2012) concluded that model self-congruity (i.e., similarity between the avatar and the user) increases the likelihood of shopping on the website. Yang and Xiong (2019) found that VFRs with personalized avatars generally outperform nonpersonalized VFRs. In contrast, our findings in Study 4 indicate that a completely personalized avatar (i.e., a virtual model that mirrors both the user's body and face) may not yield the optimal outcome. Instead, avatars with a lower degree of personalization (e.g., with users’ own body shapes but not their own faces) enhance VFR effectiveness for high-BMI users, because they are less likely to identify with such avatars and thus experience lower self-image threat. Hence, our findings add new and more nuanced insights into the role of personalization in VFRs.
Implications for Practitioners
Because we document the causal effects of VFRs on product outcomes and incorporate real-world transactional data with actual purchases, our results can be readily applied by marketers to improve sales performance and to enhance customers’ psychological well-being. Marketers should be cognizant of the diverging impact of VFRs on consumers of varying body sizes. Overweight consumers constitute a large market segment but have traditionally been overlooked by fashion retailers and understudied in the literature (Fryar, Carroll, and Afful 2020). The overweight population tends to experience anxiety and depression as psychological consequences of the prevalent weight stigma (Puhl and Heuer 2009). As a case in point, Lizzo, the Grammy-winning singer who has made significant efforts with her music to advocate overweight acceptance and body confidence, spoke about her own ongoing struggles with body dissatisfaction as “negative thoughts about her body” are triggered in her mind time after time (Aderoju 2020). We show that, when VFRs highlight their own body image, high-BMI customers experience self-image threat due to the stigma associated with being overweight, thus providing lower evaluations and exhibiting reduced purchase intentions for apparel items.
By identifying the measures to overcome such negative effects, this research presents novel insights into leveraging VFRs to benefit both businesses and customers. We suggest several ways for marketers to lessen the self-image threat that VFRs may impose on overweight customers. According to our findings, protecting customers from self-image threat not only is socially responsible but also pays off financially as it helps protect the retailer's bottom line. Online retailers should take the nature of the customer into consideration when designing and recommending the type of VFR avatar (e.g., with mannequin face or the user's own face). When launching ad campaigns, retailers with VFRs should join the growing movement to promote diversified beauty norms. They can also provide consumers with opportunities to engage in prosocial behaviors. These tactics lower the perception and impact of self-image threat and enhance retailers’ sales performance. Moreover, because overweight users of VFRs are more likely to respond positively to high-status products, marketers may use personalized promotional messages (e.g., via emails or targeted online display ads) to recommend such products to them. Doing so not only helps repair and restore these consumers’ self-image, but also enhances revenue because status products are typically higher priced. All these suggested strategies are feasible in the online retailing context thanks to precision targeting enabled by digital marketing tools.
Our findings stress the importance of segmenting customers according to their body types. Besides traditional approaches to gathering information on customer body sizes (e.g., customer input in automated size suggestion tools, customer service records), recent technological advancement is making such data even more readily available. For instance, 3DLook, a startup that is developing computerized algorithms to capture body measurements from selfies, aims to provide body analytics that “help apparel companies serve a more diverse customer base” (Lee 2020). Our research offers one of the first insights into how to leverage such information to optimize e-commerce performance.
Directions for Future Research
Our research opens opportunities for future research. First, although we focus on VFRs in the online shopping context, future research can empirically investigate how our findings might apply to offline shoppers. While physical fitting rooms are not in the scope of the current research, the literature hints at the possibility that the self-image-threat effect is likely more prominent in an online digital setting than in an offline physical setting. For instance, Kiefer, Sekaquaptewa, and Barczyk (2006) document increased feelings of threat when participants viewed a digital photo of themselves in which they appeared wider. Major, Eliezer, and Rieck (2012) find that videotaping participants elevated their anxiety and psychological distress over body images. Consumers are especially sensitive to digital images of themselves because they are perceived to be longer-lasting and may serve as a potentially permanent reminder of their body imperfections (Van Dijck 2008). In comparison, one's image in a physical mirror is transitory and short-lived. Furthermore, whereas a typical VFR highlights only the user's avatar, a mirror in a store fitting room reflects background objects in the room (e.g., bench, clothes to be tried on) in addition to the consumer. Since such background “noise” tends to decrease individuals’ attention to the focal object (Higgins, Marguc, and Scholer 2012), it stands to reason that a noise-free VFR image is likely to exert a more powerful impact on consumers’ feelings and reactions than a mirror image in a physical dressing room. With these theoretical backdrops, we focus on VFRs in this research; nonetheless, how consumer responses may differ in VFRs versus physical fitting rooms remains an empirical question to be explored.
Second, beyond the four aforementioned interventions to alleviate the negative impact of VFRs, we tested two additional moderators based on our theorizing, namely, the need for uniqueness and a self-affirmation intervention in Web Appendix C. When consumers’ need for uniqueness is heightened, they are less likely to conform to the idealized beauty standard. With a self-affirmation intervention, consumers’ global sense of self-worth is strengthened. In both situations, self-image threat is reduced and the negative VFR effect dissipates among high-BMI users. Future research can explore more moderators. For instance, consumers with a strong internal locus of control may perceive an event as primarily a result of their own abilities or actions and engage in internal attribution rather than blaming the apparel product under consideration. Moreover, although our research has examined four different coping strategies that mitigate self-image threat (direct resolution, dissociation, fluid compensation [Mandel et al. 2017]; self-affirmation [Sherman and Cohen 2006]), the literature suggests two other coping mechanisms. The symbolic self-completion mechanism suggests that self-image threat is reduced when consumers signal their strength in the self-domain that is being threatened (Mandel et al. 2017). Accordingly, when consumers can symbolically signal their physical attractiveness (e.g., by posting an attractive photo of themselves on social media), the negative VFR effect might diminish. Additionally, the escapism mechanism proposes that self-image threat is ameliorated when consumers are distracted from thinking about the threat (Mandel et al. 2017). In line with this theory, when high-BMI consumers engage in distracting behaviors (e.g., hearing loud music or eating sweets) while using VFRs, their attitudes toward the apparel item are less likely to be negatively affected. Since the sample sizes of our boundary condition studies were limited by time and resource constraints, researchers may consider increasing sample sizes in future endeavors to investigate these moderators if the situation permits.
Finally, we call for future research to explore how our findings may generalize to additional consumer groups and/or product categories. For instance, people of color are often perceived as deviating from the narrowly defined beauty standard that favors light skin. Would consumers of color (nonoverweight) feel threatened by VFR avatars that represent their own skin color? In addition, while the apparel industry by itself is an important economic sector, it may be fruitful to study the effectiveness of virtual-reality applications in other industry contexts.
Supplemental Material
sj-pdf-1-mrj-10.1177_00222437231154871 - Supplemental material for Virtual Fitting Room Effect: Moderating Role of Body Mass Index
Supplemental material, sj-pdf-1-mrj-10.1177_00222437231154871 for Virtual Fitting Room Effect: Moderating Role of Body Mass Index by Shuai Yang, Guiyang Xiong, Huifang Mao and Minghui Ma in Journal of Marketing Research
Footnotes
Associate Editor
Dhruv Grewal
Authors Contributions
The first three authors made equal contributions.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The first author acknowledges financial support from the National Natural Science Foundation of China (nos. 71972035, 71832001).
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
