Exploring consumers' virtual garment fit satisfactions to predict their actual satisfactions

Garment fit satisfaction is a very complicated phenomena due to its numerous components and has always been a problem for clothing companies since consumers often complain about poor fitting. However, this issue became more significant due to high return rates of online shopping which is often associated with the fit of the garment. Virtual try-on technology is a useful tool to minimize fit issues encountered during and after online purchasing. Virtual try-on technology enables consumers to see and evaluate how a specific cloth would actually look on their own bodies. During a virtual try-on experience, consumers are expected to evaluate the garment fit by looking at a virtual image. This image consists of a virtual cloth created by a simulation software and a virtual model that represents consumer's body. For a life-like experience, virtual model must be an accurate and realistic simulation of consumer's body which is often attained by 3D body scanning. Occasions for the virtual try-on technology to be used for online shopping have been already discussed in academia and industry during the last decade, but these expectations have not been met due to expensive and inaccessible 3D body scanners. However, 3D body scanning technology has evolved since then. Nowadays, to scan an object, a surface or a human body, is as easy as using a mobile phone or a portable scanner. Even though, it is not yet that common, using virtual try-on technology for online shopping is now around the corner. When the virtual try-on technology is fully ready for the market, it can be possible for consumers to decide whether to buy a cloth depending on their fit satisfactions of a virtual garment. This anticipation makes virtual fit satisfaction as important as real fit satisfaction. In this dissertation, factors affecting consumers' virtual fit satisfactions were heeded and a further exploration of virtual fit satisfaction notion was pursued. The research model offered by LaBat and DeLong (1990) was the keystone when exploring the factors that affect consumers' virtual fit satisfactions. According to their research, consumers' garment fit satisfactions depend on some internal factors (directly related to the cloth and the consumer) and external factors (related to the fashion trends and beauty perception of societies and fashion industry). The model offered by LaBat and DeLong cannot be directly implemented to understand the factors that affect consumers' virtual fit satisfactions since the virtual try-on conditions are considerably different than actually wearing the cloth and assessing its actual fit. Hence, for this dissertation, these internal and external factors were modified and additional factors which are only applicable for virtual context were included. These novel factors offered in this dissertation were consumers' virtual body satisfactions, virtual try-on acceptance, 3D body scanning experience and the 3D fabric properties which are derived from actual fabric properties. The relationship between these novel factors and consumers' virtual fit satisfactions were questioned and tested by an experimental setup. Moreover, the association between consumers' virtual fit satisfactions and other xx ii factors which are also present during an actual try-on scenario, such as consumers' body measurements, product liking and online shopping attitudes were also investigated. Although virtual fit satisfaction is important and influential, it is actually a tool to ensure consumer's satisfaction with the real fit of the garment. Before the virtual tryon technology, it was almost impossible to predict whether the consumer would be satisfied with the fit of the garment or to what extend he/she would be satisfied after an online purchase. In this study, consumer's virtual fit satisfaction was sought as a useful tool to predict consumer's actual fit satisfaction by incorporating internal factors (consumers' body measurements, body satisfactions and fabric properties). Ultimately, in this dissertation, a research model to predict consumers' real fit satisfactions was offered, experimentally measured, and statistically tested. Garment fit satisfaction is highly subjective since personal preferences are involved. People may like or dislike the fit of a garment due to its design that consists of its silhouette, line, color and texture (or fabric). In this study, the impact of fabric properties on consumers' fit satisfactions was particularly investigated. To do that, other design elements (silhouette, line and color) were fixed, and only fabrics have been altered. Eventually, fourteen woven fabrics were collected, and their critical properties (thickness, weight, bending rigidity, shear rigidity, elongation, drape coefficient, surface friction and roughness) were measured by various testing instruments. Afterwards, fabrics were statistically clustered into four groups according to their measured fabric properties. Following this objective clustering, seven of the fabrics were extracted due to their subjective alikeness to the other fabrics within the same cluster. Finally, seven circular skirts which have the same color (black) and same measurements (medium size), but different woven fabrics were produced by a professional seamstress. For the garment fit experiments, forty-five medium size subjects were recruited in Cornell University. Subjects' body satisfactions, product likenesses, online shopping attitudes were measured by an online survey. Then, each subject was 3D body scanned with their underwear, and their 3D body scans were exported to a commercial 3D garment simulation software to be assigned as virtual models during virtual try-on. After 3D body scanning, each subject tried on four different skirts (one skirt from each cluster) and evaluated their actual fits. Subjects also saw the simulation of each skirt before wearing and evaluated their virtual fit on their own avatar. By doing so, subjects were intended to have an ideal virtual try-on experience during an online shopping scenario. At the end of the experiments, subjects' i virtual fit technologies used throughout the experiments (3D body scanner and virtual try-on) were measured. Collected data was analyzed by using SPSS v.25 statistical software. The first objective of this research was to investigate the factors affecting the virtual fit satisfactions of the consumers. Pearson's correlation test was conducted to explore the correlation between subjects' virtual fit satisfactions and; body measurements, virtual body satisfactions, liking of the cloth design, interaction with virtual fit technologies, attitudes toward product type, and 3D fabric properties. The strongest correlation of subjects' virtual fit satisfactions was found with their virtual body satisfactions at hips level (r (180) = .50). It was followed by their belly, hip-thigh and waist measurements (r (180) = -. 34, -. 29,-.28, respectively), and finally their liking of the cloth design (r (180) = .33). A weak correlation was found between consumers' virtual fit satisfactions and their interaction with 3D body scanner and virtual try-on technology (r (180) = .26 xx iii and .24, respectively) (All correlations were significant at p=0.01 level). On the other hand, no significant correlation was found between consumers' virtual fit satisfactions and 3D fabric properties. The second objective of the study was to develop a statistical model to predict consumers' real fit satisfactions and to test its validity. In this study, various skirts were fitted on numerous subjects hence, it was appropriate to use multilevel linear regression instead of simple regression. Therefore, a multilevel linear model (MLM) with maximum likelihood estimation was developed. Subjects' real garment fit satisfaction was assigned as dependent variable; and subjects' body measurements, body satisfactions, virtual fit satisfactions, and fabric properties were assigned as predictor variables. The results of the analysis showed that subjects' virtual fit satisfaction was the strongest predictor of their real fit satisfaction (F (1, 80.72) = 55.00, p< .01, β= 0.54). The second-best predictor was fabrics' bending rigidity in weft direction (F (1, 155.26) = 13.33, p< 0.01, β = - 0.37). Subjects' waist height and overall body satisfactions were also significant to predict their real fit satisfactions (F (1, 37.67) =6.23, p< 0.01, β=0.28, F (1, 41.58) =5.71, p< 0.05, β= 0.24, respectively). Ultimatly, subjects' real fit satisfactions were predicted by the developed MLM by using their virtual fit satisfactions, body measurements, body satisfactions and fabric properties. According to the results, subjects' real fit satisfactions and their predicted fit satisfactions were strongly correlated (r (180) = .775, p < .01). Furthermore, cross validation was established to test the efficacy of the developed prediction model. For cross validation, collected data were partitioned into two groups randomly. 80 percent (N=148) of the data were used for training the model and 20 percent (N=32) were used for testing. A linear regression equation was obtained by running the developed MLM with the training data. Afterwards, obtained regression equation was used to predict subjects' real fit satisfactions in the preserved test data. According to the crossvalidation, the predicted and actually measured real fit satisfactions were strongly correlated (r(32)= .83, p< 0.01). In addition to correlation analysis, a paired t-test was conducted between subjects predicted and actually measured real fit satisfactions. According to the paired t-test results, no significant difference was found between these two groups (t(1, 31)= 1.86, d=.23, p= 0.073). The correlation and paired t-test analysis are strong indicators that the developed MLM works well to predict consumers' real fit satisfactions. Another objective of this study was to compare consumers' real and virtual fit satisfactions. The correlation between subjects' real and virtual garment fit satisfactions were explored by Pearson's correlation test and the means were compared by paired t-test. The results showed that subjects' real and virtual garment fit satisfactions were highly correlated (r (180) = .63, p< 0.01). Even though the correlation was high, a significant difference between subjects' real and virtual fit satisfactions were detected for some of the skirts particularly. This brings an answer to the final objective of the study which was to understand how different fabrics affect consumers' garment fit satisfactions in real and virtual context. The results showed that whilst consumers' grasp the fabric related fit alterations in real context, they do not perceive it in virtual context. Finally, the decline of subjects' body satisfaction level after being exposed to their 3D body scan images was also notable. This decrease was remarkable especially for waist and hips area. xx iv In this dissertation, garment simulation technology was used to create an online purchasing scenario. Moreover, consumers' garment fit satisfactions were measured with respect to the subjectivism of the garment fit whereas literature often adopt an objective approach. Factors affecting the virtual fit satisfaction of consumers were empirically explored, and a statistical model to predict consumers' real fit satisfactions were successfully developed by incorporating their virtual fit satisfactions. The research methodology used in this dissertation study can be used to predict consumers' real fit satisfactions for other cloth types or styles by adjusting variables, and consequently increase online shopping and reduce garment returns related to fit issues.