
Andrey Andreev, METRO
In this case, we provided a promo code with a discount coupon to customers identified by our AI/ML system as high-risk for churn. Our goal was to use deep learning and ML algorithms to pinpoint customers who were likely to discontinue purchases with us and offer them an exclusive discount to incentivize them to stay with Metro.
We conducted an A/B test, which demonstrated with 99.83% probability of being the best variant that offering a discount coupon increases purchases by 7.2% and helps reduce customer churn by up to 10%.
We developed and implemented a specialized AI/ML churn prediction model, trained on real-world data from Metro’s e-commerce CRM system. This model identified users with a 70% probability of churn. We then conducted an A/B test, offering a 15% discount promo code to half of these users while withholding it from the other half. The results confirmed that users who received the promo code were significantly more motivated to make a purchase.
To combat e-commerce churn, we built an AI model analyzing CRM data to predict at-risk customers (70%+ accuracy). Through A/B testing, high-risk users receiving 15% discount codes showed: -Significant conversion lift; -Improved long-term engagement; -Validated predictive model. Proving ML-driven retention works, future optimizations will personalize offers by customer value.