Size recommender: a data-driven approach to fashion sizing

Online shopping has transformed fashion retail, offering convenience but posing challenges. At the forefront of it is the persistent problem of size returns, stemming from the limitations of online shopping where customers are unable to try before they buy.

We would like to introduce a data-driven size recommender to address this dilemma and enhance not only the cost efficiency but also customer satisfaction. By leveraging historical return data, we analyzed and labeled articles, discerning size patterns, and then employed machine learning to predict size labels. The performance was validated through rigorous A/B testing, and it showed remarkable success by a promising reduction in returns.

Vorkenntnisse

A general understanding of data science, machine learning, and data-driven solutions in the context of retail would enable attendees to fully comprehend the content.

Lernziele

In this presentation, we delve into the journey of developing and implementing our data-driven size recommender. From the initial data analysis and labeling to the application of machine learning techniques for prediction, we outline a comprehensive process to improve the online shopping experience.

Speaker

 

Jie Bai
Jie Bai is a Data Scientist at the center of excellence at E.Breuninger, developing the data-driven solutions that enable to deliver an outstanding shopping experience to customers. Before joining Breuninger, she worked mainly on academic research. She has a mechanical engineering and industrial engineering background, and holds Ph.D. in Operations Management.

Jin Liu
Jin Liu is a Data Scientist at E.Breuninger. After completed M.Sc in computer science she begins her journey in the realm of data science. She enjoys discovering insights from data and believes in the potential of data to be a guiding force in making informed choices that positively impact individuals and organizations.

data2day-Newsletter

Ihr möchtet über die data2day
auf dem Laufenden gehalten werden?

 

Anmelden