What Is a Feature Store?

Feature Store Explainer

A feature store is a system for managing machine-learning features so that teams can define, discover, compute, and reuse model inputs. It aims to make the same feature logic available for historical training data and for predictions in production.

An offline store supports large historical queries used to build training datasets. An online store serves recent feature values with low latency. A registry records definitions, owners, data types, sources, and other metadata needed to understand and govern the features.

Feature stores can reduce duplicated work and training-serving skew, but they do not make features correct automatically. Point-in-time joins must avoid using future information, and freshness requirements must match how quickly the underlying behavior changes.

Teams evaluate a feature store by consistency, latency, freshness, lineage, access control, monitoring, and operational cost. Small projects may not need a dedicated platform, while organizations with many models can benefit from shared definitions and infrastructure.

A feature store manages reusable feature definitions and values so training and online prediction use consistent inputs.

Key terms include feature definition, offline store, online store, registry, point-in-time join, freshness, lineage, training-serving skew and reuse.

Historical training requires point-in-time-correct data, while live prediction may require recent values returned within milliseconds.

The feature definition should remain consistent even when the storage and retrieval paths differ.

Incorrect timestamps, stale data, weak ownership, or poorly defined features can still propagate across many models.

Lineage, validation, access controls, service targets, and deprecation rules are as important as the storage technology.

Explore more "Explainers"

Discover additional explainers across politics, science, business, technology, and other fields. Each explainer breaks down a complex idea into clear, everyday language—helping you better understand how major concepts, systems, and debates shape the world around us.