Signal Details - The models powering Raleon's AI Segmentation

The following signals use advanced AI and machine learning models to predict specific customer behaviors.

Signal Name	Description	Model Details
Ready to Buy Again	Customers that are likely to make a repeat purchase from your store (not product specific).	Uses historical order history, website activity, and email engagement data to predict repeat purchase likelihood across all SKUs; e.g., a high propensity score indicates a strong chance of repurchase.
At Risk Engaged	Customers that have engaged with your site recently, but still have a high probability of churning based on Raleon's churn prediction model.	Combines recent site visits with historical purchase and churn signals to forecast increasing churn risk despite engagement; outputs a churn likelihood score for each customer.
Promo Responsive	Customers that are less likely to purchase without a discount based on their purchase history. Discount Seekers are included in this segment.	Analyzes past purchase behavior, discount usage, and email engagement to predict discount dependency; outputs a propensity score that indicates how much a discount influences conversion.
Discount Seekers	Customers that very rarely buy without a discount. These are the most Promo Responsive.	Similar to the Promo Responsive model, but more senstive to discount dependency. Integrates Shopify order data and engagement metrics to identify customers most reliant on discounts.
Ready to Replenish	Customers who are likely to rebuy a product they have before in the next ~30 days.	Evaluates product purchase intervals and repeat buying patterns to generate a replenishment score on a per product, per customer basis; for example, a high score suggests a customer is due for repurchase the same product within an observed time window.
Very Loyal	Your most loyal customers, based on lifetime value, repeat purchase rate, average order value, and other factors.	Uses metrics like CLTV, repeat purchase rate, and AOV along with engagement data to assign a loyalty score; a high score identifies top-tier, consistently valuable customers. Your brand advocates.
On-Site Engagement (L30)	Customers that have visited your site and signaled high intent to purchase over the last 30 days.	Analyzes recent website visit frequency and specific page interactions to generate an engagement score over a 30-day period; a higher score indicates stronger immediate engagement and increased purchasing intent.
Upsell Opportunity	Customers that show potential to be loyal, but aren't quite there yet.	Merges purchase clusters, engagement metrics, and early CLTV indicators to predict upsell potential; outputs a score to identify customers ready for additional offers.
Winback Customers	Customers that were previously Very Loyal but recently show higher churn risk (may require up to 90 days after installation to fully train).	Combines historical loyalty data with recent engagement drops and churn signals to produce a winback likelihood score; helps target previously high-value customers at risk of leaving.
Refund Prone	Customers that have a high refund to order rate, and appear to have a higher likelihood to refund again.	Uses order and refund histories along with engagement data to estimate refund propensity; outputs a risk score indicating the likelihood of future refunds.
High Value Customers	Customers with the highest CLTV.	Integrates CLTV, order frequency, and AOV along with engagement metrics to score customer value with additional weighting applied to CLTV; a high score denotes top-tier customers with exceptional lifetime value.
Repeat Buyers	Customers that have purchased 2 or more times recently.	Evaluates repeat purchase behavior and recency of orders to generate a frequency score; a higher score reflects consistent buying patterns.
On-Site Engagement (L60)	Customers that have visited your site and signaled high intent to purchase over the last 60 days.	Assesses website visit patterns and behavior over a 60-day period to produce an engagement score; a higher score signals sustained purchase intent.
On-Site Engagement (L90)	Customers that have visited your site and signaled high intent to purchase over the last 90 days.	Evaluates extended site activity and browsing behavior over 90 days to generate an aggregated intent score; a high score indicates long-term engagement readiness.
Subscription Upsell	Customers that have some characteristics that indicate they might subscribe.	Combines past subscription data with current purchase and engagement trends to predict subscription potential; outputs a propensity score for subscription upsell opportunities.