Study with the several resources on Docsity
Earn points by helping other students or get them with a premium plan
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Community
Ask the community for help and clear up your study doubts
Discover the best universities in your country according to Docsity users
Free resources
Download our free guides on studying techniques, anxiety management strategies, and thesis advice from Docsity tutors
This research paper explores the impact of augmented reality (ar) on product sales, specifically focusing on how ar can reduce product fit uncertainty and increase purchase confidence. The authors examine the relationship between ar usage and product sales, considering factors such as brand popularity, product appeal, rating, and price. They use real-world data to analyze the effects of ar on customer behavior and provide insights into the potential benefits of ar for retailers.
Typology: Papers
1 / 71
This page cannot be seen from the preview
Don't miss anything!
Journal: Journal of Marketing
Manuscript ID JM.19.0703.R
Manuscript Type: Special Issue Revised Submission
Research Topics: Digital Marketing - Customer, Online Retailing
Methods: Regression Models, Logit/Probit/Tobit, Quasi-Experiment/NaturalExperiment and Diff-in-Diff
DOI: 10.1177/
Augmented Reality in Retail and its Impact on Sales
Abstract
The rise of Augmented Reality (AR) technology presents marketers with promising
opportunities to engage customers and transform their brand experience. While firms are
keen to invest in AR, research documenting its tangible impact in real-world contexts is
sparse. In this article, the authors outline four broad uses of the technology in retail settings.
Next, they focus specifically on the use of AR to facilitate product evaluation prior to
purchase, and empirically investigate its impact on sales in online retail. Using data obtained
from an international cosmetics retailer, they find that AR usage on the retailer’s mobile app
is associated with higher sales for brands that are less popular, products with narrower appeal,
and products that are more expensive. In addition, the effect of AR is stronger for customers
who are new to the online channel or product category, suggesting that the sales increase is
coming from online channel adoption and category expansion. These findings provide
converging evidence that AR is most effective when product-related uncertainty is high,
demonstrating the technology’s potential to increase sales by reducing uncertainty and
instilling purchase confidence. To encourage more impactful research in this area, the authors
conclude with a research agenda for AR in marketing.
Keywords: Augmented Reality, online retail, mobile app, virtual product experience, product
uncertainty
Despite the keen interest in AR, there has been limited research demonstrating its
tangible impact in real-world contexts. Understanding the potential for AR to increase
revenues is important in order to justify investments in this new technology. However, the
impact of AR on actual product sales is still ambiguous. By helping customers visualize
products in their consumption contexts, AR could reduce product fit uncertainty, resulting in
more sales. Conversely, AR may also discourage purchases if it leads to perceptions that the
products may not fit well. As the technology is unable to convey experiential product
attributes that could be important in purchase decisions (e.g., product texture or scent), the
impact of AR on sales could also be insignificant. This uncertainty surrounding the impact of
AR has been cited as one of the main reasons why companies are still hesitant to embrace the
technology, even though most recognize the exciting opportunities it offers (BCG 2018).
Echoing this lack of clarity, a recent article regarding applications of AR in the cosmetics
industry expressed that “Virtual lipsticks and smokey eye shadows are popular in apps, but
are they translating into more makeup sales? Hard data isn't easy to come by” (CNN 2019).
Furthermore, whether and how the impact of AR varies across different products or
customer segments is also unclear. Having a more nuanced understanding of how AR affects
sales would help marketing managers determine when it would be most appropriate to deploy
the technology. Conceivably, if AR increases sales by reducing uncertainty, its impact may
depend on product and customer characteristics that influence uncertainty in purchase
decisions, such as brand popularity, product appeal, and customers’ familiarity with the retail
channel or category. Accordingly, the present research adopts the retailers’ perspective to
examine the following questions:
How does the use of AR to facilitate product evaluation impact product sales?
How does the sales impact of AR usage differ across product characteristics, such as
brand popularity, product appeal, rating, and price?
influence the sales impact of AR usage?
Given that AR is predominantly available on mobile apps (eMarketer 2020), we focus
on the mobile app platform for our analyses. We obtained data from an international
cosmetics retailer who incorporated AR into their mobile app to help customers realistically
visualize how they look with different cosmetic products (e.g., eyeshadows, lipsticks). The
data contain sales records for 2,300 products, as well as browsing and purchase histories for
160,400 customers, allowing us to investigate how the sales impact of AR varies by product
and customer characteristics. In addition, introduction of the AR feature for two product
categories during the observation period provided us with a quasi-experimental setting to
examine the impact of AR introduction on category sales.
Findings from our research provide preliminary evidence that AR usage has a positive
impact on product sales. The overall impact appears to be small, but certain products are
more likely to benefit from the technology than others. In particular, the impact of AR is
stronger for brands that are less popular and products with narrower appeal, suggesting that
AR could level the playing field for niche brands or products at the long tail of the sales
distribution. The increase in sales is also greater for products that are more expensive,
indicating that AR could increase overall revenues for retailers. Additionally, customers who
are new to the online channel or product category are more likely to purchase after using AR,
suggesting that AR has the potential to promote online channel adoption and category
expansion. These findings provide converging evidence that AR is most effective when
product-related uncertainty is high, implying that uncertainty reduction could be a possible
mechanism for AR to improve sales.
The present research is one of the first to empirically demonstrate the impact of AR
on sales and how it varies across product and customer characteristics using real-world data.
immerses users in a completely digital environment - users are virtually transported to an
artificial, simulated world, and are entirely shut out of their surroundings. Due to the
disorienting experience of being entirely isolated from the real world and the expensive
headsets required (Ericsson 2017), the appeal of VR has largely been limited to industries
with products high in simulated content, such as gaming and entertainment (Forbes 2018). In
contrast, AR allows users to experience figments of virtual elements without the vulnerability
of being blind to the real world. In addition, AR can be experienced directly from handheld
devices that users already own (e.g., tablets or smartphones). Thus, AR is rapidly gaining
prominence and by 2022, close to 100 million US consumers are expected to use the
technology regularly (eMarketer 2020).
Augmented Reality in Retail
The unique capabilities of AR present marketers with new opportunities to engage
customers and transform the brand experience. Based on an extensive review of current
applications of AR, we identified four broad uses of the technology in retail settings – to
entertain and educate customers, help them evaluate product fit, and enhance the post-
purchase consumption experience. These uses loosely correspond to customers’ journey from
awareness to interest, consideration, purchase, and consumption, and may not be mutually
exclusive. We elaborate on the four uses below, and provide a summary with relevant
examples in Table 1^1.
-----Insert Table 1 here-----
Entertain. AR’s ability to transform static objects into interactive and animated 3-
dimensional objects offers new ways for marketers to create fresh experiences to captivate
and entertain customers. Besides generating hype and interest, marketers have also used AR-
(^1) URL links to these examples are provided in Web Appendix B.
enabled experiences to drive traffic to their physical locations. For example, Walmart
collaborated with media companies such as DC Comics and Marvel to bring exclusive
superhero-themed AR experiences to their stores by placing special thematic displays around
selected outlets. In addition to creating novel and engaging experiences for customers, it also
encouraged them to explore different areas within the stores.
Educate. Due to its interactive and immersive format, AR is also an effective medium
to deliver content and information to customers. For instance, to help customers better
appreciate their new car models, Toyota and Hyundai have utilized AR to demonstrate key
features and innovative technologies in a vivid and visually appealing manner. AR can also
be used to help customers navigate in retail stores, or highlight relevant product information
to influence in-store purchase decisions. Retailers such as Walgreen and Lowe’s have
developed in-store navigation apps that overlay directional signals onto a live view of the
path in front of users to guide them to product locations, and notify them if there are special
promotions along the way.
Evaluate. By retaining the physical environment as a backdrop to virtual elements,
AR also helps users visualize how products would appear in their actual consumption
contexts, allowing them to more accurately assess product fit prior to purchase. For example,
Ikea’s Place app uses AR to give customers a preview of different furniture pieces in their
homes by overlaying true-to-scale, 3-dimensional models of products onto a live view of the
room. Customers can easily determine if the products fit in a given space without the hassle
of taking any measurements. Fashion retailers Uniqlo and Topshop have also deployed the
same technology in their physical stores, offering customers greater convenience by reducing
the need to change in and out of different outfits. An added advantage of AR is its ability to
accommodate a wide assortment of products. By replacing tangible product displays with
lifelike virtual previews of products, retailers can overcome the constraints of physical space
Product Uncertainty in Online Retail
As consequences of purchase decisions cannot be perfectly predicted by customers,
uncertainty is inherent in market exchanges (Bauer 1960). However, it is especially
pronounced in online environments due to the spatial separation between buyers and sellers,
and temporal separation between payment and product fulfillment (Burke 2002; Pavlou,
Liang, and Xue 2007). Unlike traditional retail, customers are unable to physically inspect or
evaluate products before making a purchase, resulting in greater uncertainty that the products
would be able to deliver the expected level of performance or benefits (Bell, Gallino, and
Moreno 2018; Dimoka, Hong, and Pavlou 2012; Kim and Krishnan 2015).
Researchers have broadly distinguished between two types of product uncertainty in
online markets. Product performance uncertainty occurs when customers are unable to
evaluate or predict product performance due to imperfect knowledge (Dimoka, Hong, and
Pavlou 2012). In contrast, product fit uncertainty occurs when customers are unable to
determine if the product matches their needs (Bell, Gallino, and Moreno 2018; Hong and
Pavlou 2014). The latter form of uncertainty is typically higher for products with experience
attributes (i.e., attributes that can only be evaluated after the product has been experienced,
Hong and Pavlou 2014), such as apparel or beauty products.
Several mechanisms to reduce product performance uncertainty in online retail have
been suggested. For example, retailers could lower information asymmetry by providing
diagnostic product descriptions, or include credibility signals such as third-party product
assurances, warranties, or customer reviews (Dimoka, Hong, and Pavlou 2012; Weathers,
Sharma, and Wood 2007). On the contrary, product fit uncertainty typically requires direct
product experience to resolve, as it is idiosyncratic in nature and varies from individual to
individual. While some retailers have adopted try-before-you-buy programs (e.g., Warby
Parker’s home try-on program, Bell, Gallino, and Moreno 2018) or lenient product return
policies (Gu and Tayi 2015; Wood 2001) to provide opportunities for direct product
experiences, these measures are notoriously costly for retailers due to the additional shipping
and handling costs, and risks of product damage (Financial Times 2019). Furthermore, direct
product experiences may not be viable or appropriate in certain cases, for example, if the
product is customized (e.g., engagement rings), related to personal care (e.g., cosmetic
products), or requires assembly (e.g., furniture).
Augmented Reality and Product Uncertainty
The introduction of AR opens the possibility of substituting direct product
experiences with virtual product experiences to facilitate product evaluation and reduce
product fit uncertainty. Using a situated cognition perspective, Hilken et al. (2017) proposed
that the value of AR lies in its ability to help customers visually integrate virtual products into
the real-world environment (i.e., “environmental embedding”), and use bodily movements
and physical actions to control how products are presented (i.e., “simulated physical
control”). The unique combination of these two properties induces perceptions that the virtual
products are physically present in the real world, creating realistic product experiences.
Consequently, customers are able to evaluate products as if they are interacting with the
actual products, reducing product fit uncertainty as a result. In line with this, prior research
has found that vivid images and greater control over the presentation of information are
effective ways to alleviate uncertainty in online environments (Weathers, Sharma, and Wood
2007). By helping customers visualize products in their consumption contexts and reducing
product fit uncertainty, AR-enabled product experiences increase the level of ease customers
feel in the decision-making process, translating to positive behavioral intentions (Heller et al.
2019a; Hilken et al. 2017).
However, while AR communicates visual information about products, it is unable to
convey other experiential product attributes (e.g., product texture or scent). For example,
and customer characteristics in the following sections. Our conceptual framework is
presented in Figure 1.
-----Insert Figure 1 here-----
Moderating Effects of Product Characteristics
Brand popularity. Prior research has shown that consumers are more cautious when
they purchase from brands that are less well-known, as they anticipate feeling more regret if
the product turns out to be inferior (Simonson 1992). Consistent with this, Erdem, Swait, and
Valenzuela (2006) found that cultures that are high on uncertainty-avoidance place greater
emphasis on brand credibility. In online environments, brand signals are even more important
because consumers are not able to inspect products before purchasing (Danaher, Wilson, and
Davis 2003). However, Hollenbeck (2018) demonstrated that when additional information is
available to facilitate decision-making, consumers rely less on brand signals. As a result, less-
established brands benefit more from the increased availability of information. In the same
vein, by communicating visual information to help customers assess product fit, AR may
reduce uncertainty in online purchase decisions. Consequently, AR may decrease customers’
reliance on brand signals and inadvertently increase preference for brands that are less
popular. We use the term “popular” in a general sense to refer to brands that are more widely
adopted. Hence, we hypothesize that
H2a: The impact of AR usage on sales will be stronger for brands that are less popular
Product appeal. Within the same category or brand, products may also have different
levels of appeal due to the alignment between their inherent characteristics and general
consumer preferences. For example, a red lipstick is more mainstream and has broader appeal
compared to a blue lipstick. We draw a distinction between brand popularity and product
appeal – the latter depends on intrinsic properties of the product and could be independent of
the brand. Thus, a red lipstick from an unknown brand could have broad appeal but low brand
popularity, while a blue lipstick from a well-known brand could have limited appeal despite
having high brand popularity. As products with broad appeal cater to the masses, they are
more likely to match the needs of the general consumer. Conversely, since products with
narrower appeal (sometimes referred to as products at the “long tail” of the product sales
distribution, e.g., Brynjolfsson, Hu, and Simester 2011) serve a niche segment, there is a
higher probability that they will not match the preferences of the general consumer and thus,
carry greater product fit uncertainty. Nevertheless, Brynjolfsson, Hu, and Simester (2011)
demonstrated that in online contexts, search and discovery features, such as search tools or
recommendation engines, can shift consumers’ preferences to niche products by lowering the
cost of acquiring product information. Consistent with this, Tucker and Zhang (2011) found
that products with narrower appeal benefit more from greater information availability. By
visually conveying product information to help customers assess product fit in an effortless
and risk-free environment, AR could similarly have a stronger impact for products with
narrower appeal due to the higher product fit uncertainty associated with these products.
Therefore, we hypothesize that
H2b: The impact of AR usage on sales will be stronger for products with narrower
appeal
Ratings. Customers often turn to online ratings or reviews as a source of information
to resolve uncertainty about product quality and fit (Chen and Xie 2008). In line with this,
Kübler et al. (2018) found that consumers from countries that are high on uncertainty
avoidance are more sensitive to both the valence and volume of product ratings. However, as
consumers tend to overrate direct experiences with products (Hoch 2002), the ability to
evaluate products and resolve uncertainty via first-hand experiences on AR may reduce
customers’ reliance on online ratings. Thus, by enabling customers to learn about product
benefits and assess product fit through their own virtual experiences, AR could diminish the
retailer’s online channel (but have made prior purchases at the retailer’s offline channel) are
not accustomed to making purchases in the absence of actual products. As a result, they may
experience greater product fit uncertainty, and may be deterred from purchasing online due to
the inability to assess product fit. Since AR simulates the in-store experience of trying
products, it may help to reduce product fit uncertainty for customers who are new to the
online channel. These customers may derive greater value from the ability to evaluate
products virtually, and could be more likely to purchase online after using AR. Hence, we
predict that
H3a: The impact of AR usage on sales will be stronger for customers who are new to
the retailer’s online channel.
Category experience. Besides channel experience, customers’ familiarity with the
product category also affects their level of product fit uncertainty (Hong and Pavlou 2014).
Customers who are familiar with a product category can draw on their prior experiences as an
information source to form judgements about products (Smith and Swinyard 1982). As a
result, they may rely less on AR in their purchase decisions. Conversely, customers who are
unfamiliar with a product category lack the necessary category knowledge to evaluate
product attributes and at the same time, may not be aware of their own preferences (Hong and
Pavlou 2014). Consequently, they will have more difficulty assessing if a product’s attributes
match their preferences, resulting in greater product fit uncertainty. By helping customers
visualize how products would appear in their actual consumption contexts, AR could reduce
product fit uncertainty and increase purchase confidence for customers who are new to the
product category. As a result, AR usage may have a stronger impact on the purchase
decisions for these customers. Therefore, we predict that
H3b: The impact of AR usage on sales will be stronger for customers who are new to
the product category.
To summarize, we propose that AR usage will positively impact sales by reducing
product uncertainty. Following this line of reasoning, we developed several predictions about
which products are more likely to benefit from AR, and which customers are more likely to
respond to AR.
Empirical Analysis
Empirical Context
As AR is predominantly available on mobile apps (eMarketer 2020), we focus our
analyses on the mobile app platform. To test our hypotheses, we obtained data from an
international cosmetics retailer with both online and offline presence. Leveraging AR
technology, the retailer integrated a new feature on their existing mobile app that allows
customers to virtually try on make-up products (e.g., eyeshadows, lipsticks). The AR
technology detects customers’ facial features via their smartphone cameras and super-
imposes the shade of chosen products onto a live view of their face in real-time, giving them
a realistic visual representation of their appearances when they use the products. The brand,
product name, and price are displayed at the top of the screen. Figure A3 in Web Appendix A
provides a visual example of a customer trying on a lipstick using the AR feature. For
comparison, the corresponding product detail page view (i.e., the conventional way of
conveying product-related information on mobile retail apps) is also provided. Prior to the
start of our observation period in December 2017, the AR feature was only available for lip
categories (i.e., lipstick and lip gloss), and was subsequently introduced for eye categories
(i.e., eyeshadow and eyeliner) in March 2018. Figure A4 in Web Appendix A provides a
visual overview of AR availability for the different categories.
to examine how customers’ channel and category experience (prior to the introduction)
moderate the impact of AR usage on purchase probability.
Product-level Analysis
As product color is an important factor in cosmetic purchases, we consider each
shade/color of retail merchandise as a unique product. In total, we have 2,334 products in the
lipstick and lip gloss categories (1,984 products across 41 brands for lipstick; 350 products
across 28 brands for lip gloss). Our empirical strategy is to relate the number of customers
trying each product on AR during a particular time period with sales volume for the same
time period. We estimated the model at the monthly-product level, giving us a total of 44,
observations (2,334 products × 19 months from Dec 2017 to June 2019). As one of our
objectives is to examine the moderating effect of product ratings, we included products with a
rating in the main analysis, and replicated the analysis for all products as a robustness check.
Our final sample for the main analysis consists of 29,345 observations.
Model specification. For each product i, we model how the volume of AR usage in
month t, AR Usageit, influences the number of products sold in month t, Product Salesit. As
Product Salesit is a count variable with significant over-dispersion (M = 0.46, SD = 1.73), and
over 80% of observations are “0”, we use a zero-inflated negative binomial model for the
estimation. The vector of covariates in the regression is given by the following equation:
(1) 𝐗𝐢𝐭𝛃 = β 0 + β 1 AR Usageit + β 2 Brand Popularityit + β 3 Appealit + β 4 Ratingit + β 5 Priceit
T ― 1
m = 1
In Equation (1), AR Usageit is measured as the number of customers using AR to try
product i during month t. As brands that are more widely adopted should have higher sales,
and since the web and app channels are both online in nature and carry identical products, we
use total brand sales (within the category) from the web channel during the same period as a
proxy for brand popularity, Brand Popularityit. Following prior research using product sales
as an indicator of mass or niche appeal (e.g., Brynjolfsson, Hu, and Simester 2011), we use
total product sales from the web channel during the same period to reflect product i’s breadth
of appeal, Appealit. Ratingit and Priceit are the rating and price of product i at time t,
respectively. To examine how the impact of AR is influenced by brand popularity, product
appeal, rating, and price, we included the corresponding interactions in the model.
Additionally, we included Categoryi (1 = lipstick, 0 = lip gloss) and a series of dummy
variables, Montht, (for t = 1,…,T months) to control for category and month effects. Table 3
provides a summary of how the variables are operationalized and their descriptive statistics,
while their correlations are provided in Web Appendix C. All the correlations are low and the
variance inflation factors (VIF) are below 1.62, indicating that multi-collinearity is not an
issue. To prevent overestimation of effects due to the panel structure of the data, standard
errors are clustered at the product level (e.g., Tucker 2014).
-------Insert Table 3 here-------
Identification strategy. Our objective is to understand how the volume of AR usage
for product i during month t, AR Usageit, influences product sales, Product Salesit. However,
AR Usageit could be endogeneous, as customers may be more inclined to use AR to try
products that they are interested in purchasing. To account for this endogeneity, we use the
two-stage residual inclusion method (Terza, Basu, and Rathouz 2008), which has been used
in recent research when both the endogenous and dependent variables are non-linear (e.g.,
Arora, ter Hofstede, and Mahajan 2017; Danaher et al. 2020).
Following the two-stage residual inclusion method, we first regress the endogeneous
variable, AR Usageit, on all other covariates in Equation (1). Residuals from this first stage
