User Intent Recognition Framework

Our intent structure consists of four layers:

① Core Interests: Describes the "keyword cloud" in the user's mind, their psychological profile while browsing the store.
② Product Focus: Defines from a product perspective "what is the core appeal of this product".
③ Purchase Signals: Reflects which layer of the "purchase funnel" the user is currently in.
④ Behavior Summary: Converts all behaviors into one sentence "how the user is browsing".

These four layers together constitute "our understanding of user psychology and behavior".

All user behaviors can be categorized into three types of signals:

① Exploration Signals (e.g., staying on homepage, browsing category pages) → User is still looking at "overall style".
② Comparison Signals (e.g., repeatedly viewing a product, focusing on size/price) → User is filtering options.
③ Decision Signals (e.g., adding to cart, checking inventory) → User is ready to purchase or hesitating.

In this example, the user mainly shows exploration signals, so we judge them to be in a weak intent, early stage.

Based on user behavior, we execute three steps:

① Capture Behavior → ② Extract Signals → ③ Generate Intent

You can see a visual version of this process below.