Intent and propensity
Intent is what a customer is doing right now; a propensity model predicts what they are likely to do, learned from past behavior. These explainers cover what propensity modeling and scores are, how intent read live differs from a predicted score, why sorting a visitor into a finite set of buckets is a different problem from ranking a catalog, and why propensity models so often stall before they ship.
In this topic
What is propensity modeling?
Propensity modeling is the practice of predicting how likely a customer is to take a specific action, such as buying, churning or converting. A model learns the pattern from labeled history and outputs a probability for each customer.
What is a propensity score?
A propensity score is the number a propensity model produces: a probability between zero and one that a given customer takes a given action. It ranks customers by likelihood, but it does not say why any one of them scored the way they did.
What is real-time intent?
Real-time intent is what a customer is trying to do right now, read from what they are doing in this session. It is observed from current behavior rather than predicted from history, so it changes as the session changes.
Intent vs propensity
Intent is what a customer is doing in this session; propensity is a model's prediction of how likely they are to take an action, learned from history. One is observed and current, the other is predicted and trained, and they answer different questions.
Propensity score vs intent bucket
A propensity score represents a customer as a continuous probability. An intent bucket represents them as one named state from a small, finite set, like browsing, comparing or ready. The two are different ways to carry the same decision, and they behave differently.
Classification vs ranking
Classification sorts something into one of a finite set of categories. Ranking orders a large set of items by a score. Reading a visitor's intent is classification into a handful of states; choosing items from a catalog is ranking. They are different problems with different right tools.
Why propensity models stall in production
Propensity models often stall before they change what a customer sees. The gap is between a working notebook and production, where labeled data, a feature pipeline, offline validation, a serving path and train-serve skew each have to be solved before the first score ships.
Start with the free tier
Read the ideas, then put them to work: add the SDK, define one attribute, and see an attribute updating against your own traffic in the first session. The free tier runs on the real engine: no card, no sales call.