Hydrant triples revenue per customer with Pecan AI predictive winback in 2 weeks
Hydrant, a wellness DTC brand, used Pecan's predictive AI to identify churned customers most likely to return — achieving a 260% higher conversion rate and 310% more revenue per customer versus control.
Background
Hydrant heavily relied on email marketing but lacked the ability to target offers based on individual customer behavior or purchase likelihood. Broad segments led to wasted marketing spend — discounts sent to customers who would have returned anyway, and insufficient outreach to those who needed incentives.
What Was Implemented
- Pecan AI predictive analytics platform integrated with Klaviyo, Shopify, and Snowflake
- Churn prediction model built on 180-day customer purchase history
- Individual-level scoring for: churn likelihood, subscription conversion potential, and winback probability
- A/B tested winback campaign targeting cohort predicted least likely to return organically
- Model built in approximately 2 weeks
Results
In A/B testing, customers predicted by the Pecan model to have the lowest chance of purchasing again — when targeted with offers — achieved a 260% higher conversion rate and 310% higher average revenue per customer than the control group. Of customers predicted to be least likely to churn, fewer than 2% actually churned; of those predicted to be most likely to churn, over 83% did churn — confirming strong model accuracy.
Lessons
- Targeting lapsed customers who need incentives (rather than all lapsed customers) dramatically improves campaign efficiency
- Predictive models can be built rapidly on existing commerce data stacks without a dedicated data science team
- A/B testing the model against a control group provides a rigorous proof of lift, not just observed improvement