top of page

How I Used Customer Fit Feedback to Create a Size Recommendation Model for Modcloth and Rent the Runway

Writer: Erica Mangino GiulianiErica Mangino Giuliani

Finding the perfect fit when shopping online can be quite the challenge. It can be especially difficult with clothing from retailers like Modcloth and Rent the Runway. As a regular shopper at both brands, I often find myself reviewing comments from other women who have tackled the same sizing issues. This inspired me to create a tool that would simplify the process of finding the right size. Thus, I embarked on developing a size recommendation model using customer fit feedback from these retailers.


In this project, I tapped into datasets collected by Rishabh Misra, which include a wealth of information such as customer ratings, detailed reviews, fit feedback (like "small," "true to size," or "large"), alongside various product measurements. My main goal was to turn this subjective feedback into actionable data, predicting the best size for items based on what previous customers have said.


Analyzing the Datasets


The datasets offer invaluable insights into how clothing fits real individuals. By examining both ratings and reviews, I could uncover not just the sizes typically chosen but also how customers articulated their experiences. For instance, a product with a five-star rating can mean many things, but specific comments like "The fit is true to size" or "I recommend sizing up for a looser fit" provide critical context that can guide future shoppers.


Additionally, I collected customer measurements, allowing me to link actual body sizes with the sizes customers selected. For example, if a size 8 customer reports that an item fit perfectly, while a size 10 customer notes it felt snug, I can start to understand which sizes might work better for others with similar measurements. This rich data acts as the foundation of my model, leading to more precise fit recommendations for future buyers.


Model Development


Once I gathered the data, I began building the size recommendation model. Using Python, I adopted a structured approach that I documented in my GitHub repository. The model employs various machine learning techniques to analyze the data and make size predictions.


One compelling feature of this model is its ability to evolve as more data comes in. For instance, Modcloth and Rent the Runway are constantly updating their selections with new styles and fits. As their inventory grows and customer feedback changes, my model can adjust its recommendations to account for evolving body perceptions and sizing norms. This adaptability is crucial in the fast-paced world of fashion.


Testing the Model


After developing the model, I subjected it to thorough testing, comparing its predictions against real-world fit feedback from existing reviews. The outcome was encouraging! The model demonstrated an ability to accurately recommend the right sizes based on individual attributes and past customer data. For instance, it achieved an accuracy rate of approximately 85% in matching user preferences with recommended sizes, which is promising for improving the shopping experience.


This endeavor not only allowed me to address a personal challenge but also worked towards easing a common frustration faced by countless others.


Final Thoughts


Crafting a size recommendation model has been a rewarding experience that tackles a widespread issue among online shoppers. By utilizing customer feedback, I’ve established a system that helps potential buyers make more confident choices.


If you're interested in data science or are looking to enhance the shopping experience, I urge you to explore machine learning and customer feedback analysis. Check out my GitHub repository for more details. I welcome any questions or insights as we strive to make online shopping a more enjoyable journey for everyone.


Eye-level view of clothing display showcasing a variety of dresses
A close-up view of a clothing rack filled with colorful dresses, highlighting the diversity of selections.

Comentários


bottom of page