Teams across Supertails were personalizing blind, with no shared read on what each customer actually cared about.
A scoring system that turns raw behaviour into a normalized affinity for every customer, across pet type, category, sub-category, and brand.
Teams across Supertails were personalizing blind, with no shared read on what each customer actually cared about.
Behaviour is not preference. The win was turning raw events into a weighted, normalized score for what each customer actually cares about, at every level of the catalog.
I built a multi-level affinity layer scoring preference across pet type, category, sub-category, and brand, time-decayed and intent-weighted.
Personalization is a problem of understanding the customer before it is a model problem. Get the customer read right and every team downstream benefits.
Open the full architecture, modeling, validation and lessons.