8% profitability boost through personalization using multi armed bandit techniques
Used Multi Armed Bandit (MAB) Reinforcement Learning techniques to create personalized strategies to boost the client’s profitability and improve their average sales per consultant by over 24% as compared to their earlier system.
The client is a major cosmetics company and wanted to develop a personalized digital recommender system for their recently built website to maximize the average sales per order, conversion rate and customer life-time value. The existing system had a fixed set of business rules to determine the display rank of offers and didn’t have ways to test and understand the performance of personalization strategies.
Sigmoid built an end-to-end system for generating personalized recommendations (for digital subscribers) based on customers’ historical purchase history using multi armed bandit recommendation system. The multi armed bandit solution was implemented to learn from personalized recommendation strategies and select the best performing strategies for a specific user for future campaigns, based on real-campaign feedback for consultants.
Implemented over 100 strategies across 8 countries using multi armed bandit (MAB) based solutions. The resulting impacts were 8% improvement in profitability and 24% improvement in average sales per consultant.