Review of Data and User Interface in Online Apparel Size Recommendation Systems

Kim., Y. E. & Choi., H. E. (2025). Review of Data and user interface in Online Apparel Size Recommendation Systems. Journal of the Korean Society of Clothing and Textiles, 49(2), 265-281. https://doi.org/10.5850/jksct.2025.49.2.265

As online shopping becomes increasingly integral to our daily lives, one of the biggest challenges in apparel purchasing remains size selection. While the online fashion market reached approximately 37 trillion KRW in 2024, accounting for over 33% of the total online market, clothing items represent 31% of all e-commerce returns, with 45% of these returns attributed to issues with size, fit, and color.

To address this persistent challenge, size recommendation systems have emerged as a promising solution. This study analyzes 15 currently operating size recommendation systems to identify their characteristics and propose directions for improvement.

Current size recommendation systems can be categorized into three main approaches based on their data collection methods:

1. User Input-Based Collection

This approach relies on users directly inputting their body information through surveys with questions like "Are your shoulders broad?" or "How would you describe your waist?" Users can also manually enter their measurements. While convenient, this method is limited by subjective judgment and potential inaccuracies in self-perception.

2. Image-Based Measurement Extraction

This method uses smartphone cameras to capture full-body photos, which AI algorithms then analyze to extract body measurements. Most systems require two photos (front and side views), with shooting distances varying from 1.5m to 3-4 steps back depending on the system. While this provides more objective data, accuracy can vary based on shooting environment and methodology.

3. Product Data Integration

This approach involves measuring actual owned garments or utilizing size information from previously purchased items from other brands. Though potentially the most accurate, it requires users to perform measurements themselves, which can be cumbersome.

The methods for presenting size recommendations based on collected data fall into three distinct categories:

1. Text-Based Presentation

Simple, straightforward recommendations such as "Size M is suitable for you." While quick and easy to understand, this approach lacks detailed explanations for the recommendation rationale, which may create uncertainty for some users.

2. Visual-Based Presentation

This includes color-coded size charts highlighting recommended sizes and avatar-based fit simulations showing how garments would fit on different body parts. Some systems provide 2D illustrations or 3D virtual fitting experiences, offering intuitive and easily comprehensible information that enhances user confidence in their selection.

3. Statistical-Based Presentation

Data-driven recommendations that leverage user feedback and purchase history, such as "90% of users with similar body types chose size L" or "Based on thousands of similar shoppers, there's a 68% chance you'll be happy with size 8." This approach builds user trust through transparency but requires substantial data accumulation and well-designed feedback systems.

Based on our research findings, we propose the following enhancement strategies:

Clear Measurement Guidelines

Most systems fail to provide adequate measurement instructions, limiting user understanding. Detailed measurement protocols combining images and text descriptions are essential for improved accuracy.

Expanded Objective Data Utilization

To overcome the limitations of subjective body perception, systems should increasingly adopt objective methods such as image-based measurements.

Enhanced Visual Feedback

Intuitive and easily understandable information presentation, including 3D virtual fitting for fit confirmation, is crucial for user confidence and satisfaction.

Global fashion companies are actively adopting image-based size recommendation technologies, providing consumers with more intuitive and reliable recommendation experiences. Similar developments are expected to accelerate in the domestic market, with increasingly sophisticated size recommendation systems being developed and implemented.

Size recommendation systems represent more than just technological advancement—they fundamentally improve the consumer shopping experience while contributing to more sustainable consumption patterns by reducing returns. The evolution toward more accurate and user-friendly systems holds significant promise for transforming how we shop for clothing online.

As the technology continues to mature, we can expect these systems to become increasingly sophisticated, potentially incorporating advanced features like 3D body scanning, AI-powered fit prediction, and personalized style recommendations. The ultimate goal remains clear: creating a seamless online shopping experience that rivals the confidence of in-store fitting rooms while maintaining the convenience of digital commerce.

This research contributes valuable insights for companies looking to develop or improve their size recommendation systems, while also laying the groundwork for future innovations that could revolutionize online apparel retail.