Problem / Objective

All About Fashion (Abof) is the online venture of the Aditya Birla Group, the third largest business conglomerate in India. The Aditya Birla group is an established name in the fashion retail segment with over 2000 brick & mortar stores. After the acquisition of Forever 21, they further consolidated their position in the Indian fashion retail market. Abof understood that to succeed in improving customer’s experiences, it required to focus on smart technology and digital innovation across all aspects of the online business. One of the key areas was the use of content curation technology. The company's stylists collected user generated content (UGC) and trends from social media and hand curated looks. This gave the online retailer good results but the solution was not scalable as only a handful of products were available in the curated form each week.

Solution / Approach

Abof tested various kinds of recommendations based on product attributes and color. These solutions ranged from meta tags to collaborative filtering. After doing A/B tests, Abof zeroed in on Streamoid’s image-based recommendations, as the solution enabled an increase in click through rates from 3.4% to 9%. Abof introduced Streamoid’s Outfitter, an AI-powered recommendation engine that created complete looks for every product in the inventory, in real-time. The recommendation engine was trained on Abof’s styles, and was based on outfits modeled in the catalog and the internal style guide. The self-learning system learned about trends as well as the customer, hence, the recommendations were personalized. 

Outfitter is the result of several patented technologies. It combines a proprietary fashion rules set with data science technologies, including Computer Vision and Machine Learning. The key to Outfitter’s performance is in the ability to extract features from images and run these features through a rules engine to provide appropriate recommendations. These recommendations are provided in under 0.5 seconds. This self -learning system is highly scalable and can be easily integrated with other systems.

Impact / Implementation

The performance of Streamoid’s Product Recommendation system impacted several key metrics of Abof.com. The results of the solution post the implementation indicated that customers who utilized the Outfitter were 20% more likely to stay on the website after visiting, spending 21% more time, and were 60% more likely to purchase. The implementation led to a 6-times increase in conversions, 31% increase in basket size, and 38% greater revenue.

Photo by Priscilla Du Preez on Unsplash