Use Cases
Visual AI for Catalog “Cold Start” and “Out of Stock” Problems
Applicable Segment(s):
Fashion & Apparel, Speciality
Impacted Function(s):
Commerce
Solution:
DeepRecs Visual AI
The Company
An American multinational retail company operating a chain of hypermarkets, department stores, and grocery stores.
The Challenge
As a big department store, they refreshed their catalog, especially the apparel category, daily. They also run out of popular products quite often. So, they needed a strategy that solved the cold-start problem with new products and provided relevant alternatives when a product went out of stock.
The Approach
- Using Algonomy’s (now part of ADA) Image Similarity models powered by Visual AI, they were able to see immediate benefits.
- Each seed considers the entire catalog. The current system can stratify the catalog if labels are provided (eg; product_id, category_id).
- The model architecture is a variant of ResNet, which is a popular convolutional neural network architecture for transfer learning on images within the recommender systems domain.
