More than a few consumers, it is fair to say, do not have patience to change and try-on clothes in the fitting rooms (actually cabins) of fashion stores or departments before buying anything. The ‘ritual’ usually involves undressing, trying on a garment (e.g., a shirt or a pair of trousers) in the fitting room, moving out to look at a mirror in the floor’s main area — and perhaps walking around with the clothing or asking for second opinion, then going back and change to try another garment; this procedure may be repeated, twice, three or four times, and occasionally even more. The idea, therefore, that one can try-on clothes without physically changing their clothing appeals to consumers who find the traditional experience too much time-consuming, annoying, and possibly feel uncomfortable trying-on clothes in a public place. The idea of virtually trying-on clothing is getting feasible and more practical lately with the aid of advanced technologies (e.g., AR and AI), online as well as in the brick-and-mortar stores themselves. Health concerns due to the risk of contracting the coronavirus by touch (at least earlier during the pandemic) have generated even greater interest in virtual fitting/dressing rooms, and hastened the development of working solutions.
In principle (i.e., before going into some more detail), a shopper can see his or her own photo-image or an image of a self-resembling ‘model’ on a screen ‘dressed-up’. The shopper may be asked first to give some information about one’s dimensions (e.g., weight, height, waist) and appearance (e.g., skin complexion). Then, upon selecting a garment of interest, that garment is ‘dressed’ on the image representing the shopper to show how it fits and looks. A combination of garments (e.g., trousers, shirt and jacket, a blouse and skirt) may be tried-on together as a whole outfit. Garments can be replaced and combinations of different garments can be tried-on without requiring the shopper to physically wear the clothes.
Julia Dietmar, writing for Forbes (16 September 2021), suggests four main motivations for fashion retailers to launch virtual dressing rooms: (1) Convenient, safe fitting experiences (with emphasis on health implications of the pandemic); (2) Improve conversions (from shoppers to buyers) by allowing shoppers to make better informed decisions; (3) Make customers out of casual (online) browsers by creating more personalised, engaging and inclusive online/mobile dressing room experiences; and (4) Increase loyalty and retention by enabling retailers to learn more comprehensively the preferences of consumers and their habits or patterns of shopping for clothing (e.g., via AI-powered analytic tools).
A specific advantage is expected from the application of virtual fitting rooms in reducing return rates of clothes ordered online — that is, when consumers realise after the order arrived at home that the garments do not fit them perfectly or do not look according to the impression they got of the garments on the screen of the computer or smartphone (e.g., the colour does not match). Whilst consumers were staying at home, during the lockdowns imposed to combat the coronavirus, they resorted to purchasing clothing online more frequently (although the need overall for new clothing declined), and so the problem of returns also became more acute. First, it is expected that on the basis of measures of a shopper’s body it would be possible to make more accurate size recommendations. Second, it is hoped that the appearance of clothes, especially when tried on a representative image of the shopper, would be more realistic and authentic (particularly when garments chosen make-up an outfit)[see Sarah Perez, TechCrunch, 13 May 2021; Michael Waters, Modern Retail, 17 May 2021]. Additionally, applications of virtual dressing rooms are about to support social interactions with respect to the clothing items users consider buying. Early applications already offer or promise the possibility to upload the representative image of the user dressed with the garments of choice to a social media platform, seeking the opinion and advice of one’s network associates (e.g., friends and relatives). Waters (Modern Retail) suggests that social interaction can help encourage shopping, generate word-of-mouth (WOM), and accelerate adoption by more retailers.
Different approaches and solutions are being devised in order to produce a virtual dressing room application. The application has to be practical, convenient or easy-to-use, and visually compelling. We will describe here in just general terms three key approaches:
- A ‘model’ image or figure (‘avatar’) of the shopper is constructed on the basis of measures he or she reports to the application. The representative image has to maintain the proportions of body dimensions of the shopper (e.g., height, weight, waist, body-build of torso and shoulders), plus optional characteristics of appearance (e.g., skin complexion, hair colour). The garments selected by the shopper-user from the online catalogue are then ‘dressed’ (as layover) on the representative image (which may be more or less visually resembling the shopper). This approach relies mainly on artificial intelligence (AI) capabilities to construct the figurative image representing the shopper and fitting it with the images of clothing items.
- A much talked-about approach is to simulate a mirror that can dynamically add clothes to the ‘reflected’ image of the shopper. Reflection of the shopper is actually created by taking a photo-image of the shopper, which thence can be manipulated or superimposed in order to dress it with the selected garments. The application also has to be able to track movements of the shopper-user in order to show his or her image from different angles. The approach is based on technology of augmented reality (AR) — it is inspiring and imaginative, aspiring to the status of an ‘ideal’ solution. However, the AR-driven approach has confronted some difficult technological obstacles to practise it reliably and convincingly, most notably in creating 3D visual representations of the shopper and the clothing items, and fitting them together.
- A third sophisticated approach overcomes weaknesses of the AR-driven approach while creating more compelling and realistic outcomes than the AI’s approach based only on measures reported by the shopper-user. The process starts with taking a photographic image of the shopper-user. Yet, this image is used as the ‘template’ source for constructing a computer-graphic model of the shopper that will retain his or her visual appearance; most measures of the shopper’s body can also be estimated from the photo image in order to create a figure that matches the body dimensions and shape of the shopper — this is essential of course for providing more accurate size recommendations. Image representations of clothing items thereon can be fitted to and dressed on the model image resembling the shopper. This approach draws on AI capabilities.
Zeekit is an Israeli-based start-up that marries advanced technology with fashion, founded in 2013 and recently (~April-May 2021) acquired by American retailer Walmart, one of Zeekit’s large clients. Its solution for a virtual fitting room corresponds to the third approach described above. More specifically, Zeekit’s application is based on an original multi-layer imaging technology and AI Deep Learning models; the idea emerged in quite a different field, topography, where the founders called their method “dressing mountains”, and then thought of adapting it to “dressing people”. The solution of Zeekit is furthermore genuine by building graphic models of the human body and the clothes, being broken into 80K components of ‘3D nets’ and recomposed to flexibly manipulate, fit, and position the body with clothes tried-on. Clients of Zeekit included so far fashion designers / brands and retailers (e.g., Macy’s, Tommy Hilfiger, ModCloth, Walmart) that purchased a license for the application for an annual fee on the basis of software-as service (SaaS). But following the acquisition, Walmart will get the exclusive rights for deploying the technological solution and application capabilities of Zeekit in its retail activities [*].
The application of Zeekit can offer some variations for creating the image to represent the users for dressing. Primarily, the image can be based on a photo image taken by the user, complemented with a few details provided explicitly (particularly height). Yet, a shopper-user may choose not to provide one’s own photo and instead be assigned the image of a fashion model to represent her (recommended to match most closely measures reported by the user). Apparently Zeekit could offer this option because it also marketed its application to fashion houses for virtually dressing their professional models (e.g., at virtual fashion shows). It is not yet clear which features and tools of Zeekit will be used by Walmart and how, but it is already said the format is likely to be different. The shoppers at Walmart can be expected to have a choice between uploading their own photo image as ‘model’ or applying a ‘model’ (like avatar) configured to represent their reported measures, shape and skin tone (Perez, TechCrunch).
Yael Vizel, CEO and co-founder of Zeekit (together with Nir Appleboim, VP R&D, and Alon Kristal, CTO) proposed in an interview to Israeli business newspaper TheMarker [*] the key strengths in her view that Zeekit focused on developing: (1) Enabling shoppers to share dressed images of them in social media network sites (e.g., Facebook, Instagram) and conduct social interactions regarding their selections; (2) The application can be deployed and utilised both online and in-store; (3) Allowing shoppers to see and show how the clothes they choose look on fashion models representing them (note: this option may contribute towards self-image and social benefits such as making a good impression, expressing an aspiration, creating a buzz); (4) Shoppers can compose outfits that combine clothing items from different fashion brands and see how they match and look together — this is something, Vizel explains, that shoppers usually cannot do in a mall, where each store sells a specific set of brands and you cannot move items between stores before paying, but is possible in department stores or multi-category stores as of Walmart’s. In addition, Zeekit can provide a powerful knowledgebase to a retailer about the preferences, tastes in fashion and shopping patterns of consumers. Vizel believes that the technology Zeekit developed is setting a new retailing reality fit for the next generation. These strengths and capabilities have probably helped in making Zeekit attractive to Walmart.
Alternative solutions have been proposed, experimented and implemented over recent years (e.g., StyleMe — Cisco, virtual fashion mirror; Style.me — AI-based with reported measures; PICTOFiT — Creative Reality, 3D images of clothes fit to 3D ‘avatar’ of shopper; Outfit-VITO — Amazon.com, photo image-based virtual try-on system). Nonetheless, if the methodology of Zeekit becomes locked-in by Walmart it may be an unfortunate miss for the market, especially supposing its technological solution is superior. The application could be great for consumers and attractive to retailers, but the acquisition by Walmart may deprive many of them from accessing it. It might have been better if the start-up company had remained independent, by bringing in more investors, or being acquired by a larger technology company, so that the application could continue to be more widely available. Vizel indicated, however, that in reality they received more offers from players in the fashion market than investors [*] — it is seemingly more likely that, if not by Walmart, Zeekit would have been captured by another larger retailer.
Yet, even a greater problem may be looming with the acquisition by Walmart. Ronald Goodstein, professor of marketing at Georgetown University, argues in an interview to Retail Dive (Maria Monteros, 24 June 2021) that this pretty high-end technology seems to suit better upper-tier brands, which is disconnected from Walmart. As Monteros comments, the retailer is “better known for its affordable prices and not exactly for high street products.” This move is supposed to move Walmart ‘upstream’ in its pursuit to become a stronger, more serious player in fashion. As said by Goodstein regarding Walmart: “They’re basically taking [Zeekit] out of the market from everybody else”, at a time when the technology was progressing well in the market (and the pace of adoption was even exacerbated by the pandemic). If Walmart does not succeed in promoting its status with the help of Zeekit, falling short of expectations, the application could be ‘swallowed’ by the giant retailer, and that would make its disappearance a complete loss for the whole market.
Challenges remain in the face of further development and implementation of technologies of virtual fitting rooms. For example, there is a debate in the field on the quality of photographic images of shoppers required for producing representative images that are reliable, resemble the person, and functional for fitting images of clothes to them — this is particularly argued with regard to AR applications. Can any photo taken at home be sufficient or does it have to be taken in special conditions of a studio? Some already foresee applications of virtual reality (VR) with a whole scenery fabricated around the shopper trying-on clothes. While this may still be ahead of us, the requirements of VR for producing a high-quality 3D image of the shopper, free of distortions, may be similarly high for AR. Furthermore, there are also issues concerning the imagery of clothing items that seem not to be discussed enough. The usual 2D photo images used in online store websites may not be sufficient for implementing a virtual fitting room at good standard. For retailers who have thousands of items in their catalogue, taking photos of so many of them in the quality standard required may deter them from adopting the technology. The domain of clothing also has more complications than in other product categories for AR imaging: “more sizes are involved, and the technology has to be capable of estimating fit for many different types of clothes across many different body types” (Waters, Modern Retail). This issue has also been raised as a challenge by Vizel with regard to the application of Zeekit (their method of collapsing and re-composing graphic image models could be part of the answer to this challenge).
The business and trade media seem to concentrate on online shopping for clothes and do not relate enough to the possibility of applying the technology as self-service in physical stores. Amid the risks of the coronavirus, shoppers have been concerned about wearing clothes from display in a store. But even beyond that, on a more regular basis, many shoppers could benefit from greater convenience by trying clothes virtually on them, being exposed to collections in the store as well as in the online catalogue. Moreover, shoppers may ‘experiment’ first in trying-on a larger variety of items virtually, screen from them a small choice set of garments that seem most fitting, appealing and appropriate for them, and then take this final set or outfit to physically wear in the traditional dressing room of the store to conclude their purchase decision (i.e., a typical two-stage choice process consumers often follow).
There is undeniably an advantage in having a visual preview of how a desired garment looks on an image resembling the shopper (at least in measures) before shopping online away from a store. It should give a good idea if the size recommended seems appropriate as well as if the garment looks good when dressed on a body, and if it matches other garments in an outfit (e.g., a suit). It may give a good indication of fit in size-measure and appearance. But one should remember this is still a simulation of fit, and not a complete feeling of fit, for instance to feel that the trousers sit well at or below the waist, the shirt does not press on the chest or arms, the jacket stands well on the shoulders, and even if the skin is not sensitive (itching) to the fabric. For this kind of fit there is no replacement to trying-on clothes in a store.
The technology of virtual fitting rooms, in its different variations, looks promising — user applications can lead to increased convenience, confidence and practicality in shopping online. Utilising the technology in-store may also increase the efficiency, ease and comfort of shoppers in the physical stores. Virtual fitting rooms appear to alleviate some of the limitations and hurdles shoppers face when engaging in e-commerce. However, there are still issues to be resolved that may be important for retailers and consumers, impinging especially on the shopping experience, to make it more useful, flowing and visually appealing. Nevertheless, even as forms of shopping change, the incumbent and next generations may find comfort and satisfaction in combining shopping in virtual and physical fashion domains.
Ron Ventura, Ph.D. (Marketing)
[*] An article including interview with Yael Vizel, CEO of Zeekit (“The Feelings Are Complex. Zeekit Is an Organ in My Body and Now It Is Being Sold”, origin in Hebrew), Anat Georgy and Ruth Levy, TheMarker (MarkerWeek Magazine), 23 July 2021, pp. 26-28. ‘Zeekit’ means Chameleon in Hebrew.