AI Assistant (Leat MCP)
July 20
→
Propensity and affinity scoring
Understand what each customer is likely to do next and which products, categories, locations, or experiences interest them most. Propensity and affinity scores update as new activity arrives, keeping segments, automations, and decisions aligned with the customer’s latest behavior.
Propensity scores estimate the likelihood that a customer will take a particular action within a chosen period. This can include making a purchase, returning to a location, responding to an incentive, redeeming a reward, reaching the next tier, or becoming inactive. Scores update as customer activity changes, so a recent visit or an unusually long absence can immediately affect the expected outcome.
Affinity scores rank the products, categories, brands, locations, channels, and times that appear most relevant to a customer. They can reflect purchases, browsing behavior, redemptions, and preferences the customer has shared directly. A customer can show an interest in a category they have not purchased from yet, helping the program recognize possible next purchases rather than relying only on past orders.
Propensity and affinity scores become attributes on the customer profile. Use them to build segments, trigger automations, support reporting, or inform the Decision Engine. A drop in return likelihood can start a retention journey, while product affinity can help choose between relevant rewards. Score distributions can also be compared across tiers, locations, and other customer groups.
Each score includes a confidence level and the behaviors or preferences that influenced it. Teams can see why return likelihood changed or which signals created an affinity for a category. Responsible AI Controls determine which information scoring may use, preventing excluded attributes from influencing the result. A high likelihood to purchase does not automatically mean a customer needs an incentive. Uplift modeling helps determine whether taking action is likely to change what they would otherwise do.
Scored outcomes
Teams can choose which outcomes to predict, including purchase, return, response, redemption, tier progression, churn, and goals specific to the business.
Automation thresholds
Score levels can determine when an automation starts, such as when a customer’s likelihood to return falls below a chosen point.
New customer scoring
Teams can decide how customers with limited history are scored and how much influence early scores have on decisions.
Score accuracy reviews
Predicted outcomes can be compared with what customers actually did, allowing scores to be adjusted when the two begin to differ.
Score driver reviews
Teams can inspect which behaviors influence scores and confirm that excluded signals are not contributing to the result.
Catalog coverage
Products, categories, and new ranges can be added to affinity scoring as the catalog changes.
