Challenges
- Client’s human stylists usually pose questions to the customers to understand their preferences and recommend new outfits
- As the customers and catalogue size increase, it would be impossible to deliver a personalized experience by stylists
- To address the increasing demand, client wanted us to develop an intelligent AI Stylist, which can help the customers in picking outfits based on their fashion preferences, which include colors, person-size, trends and availability
Solutions
- An ML model was built using previous orders, customers’ features, demographic details, product data and user ratings to score the user
- Features such as seasonality of product and customer, affordability, style, weight, trends and personality of customer are used to learn customer preferences
- Used a hybrid filtering algorithm to provide recommendations, which uses both content and collaborative filtering techniques
- K-means Clustering is used to cluster the customers with similar features. Clustering in combination of above-mentioned filtering algorithms recommends the outfits to the customers
Tools & Technologies
Python, Sokit Learn, Angular, Prophet
Key benefits
- We were able to select the products with 90% similarity with purchased products
- Usually it takes 20 mins for a stylist to recommend an outfit accurately. This application recommends instantaneously and increased the productivity of stylists by 3-5x
- This application increased the keep rate of customers by 5-15 %
- This application allows enhance customer experience even when working with human stylists, as they can refer to historical data to understand customers preferences quickly
