Full Answer

Klaviyo's predictive analytics operate on four distinct models, each built from different data inputs and serving different marketing decisions.

Churn risk scores each customer on a scale from low to high based on their historical purchase frequency and recency. A customer who typically orders every 30 days but has not purchased in 60 days receives a high churn risk score. This enables automated win-back flows triggered by behavioral signals rather than arbitrary time delays.

Predicted lifetime value estimates total expected revenue from a customer over their relationship with the store. The model factors in average order value, purchase frequency, and historical retention patterns. Stores use this to allocate acquisition spend — knowing a customer segment's predicted LTV justifies higher cost-per-acquisition for high-value cohorts.

Expected next order date estimates when a customer will likely purchase again based on their individual purchase cadence. This prediction powers replenishment flows for consumable products and re-engagement timing for seasonal buyers.

Gender prediction uses name analysis and behavioral patterns to infer gender for personalization. This model carries lower confidence than the transactional predictions and should be used cautiously — incorrect gender assumptions damage brand trust faster than generic messaging.

All four predictions require a minimum data volume to activate. Klaviyo's models need at least 500 customers with purchase history and sufficient engagement data before predictions become statistically meaningful. Stores below this threshold receive placeholder scores that should not inform marketing decisions.