Intermediate · Data

Data Drift

Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.

TL;DR. When the statistical distribution of inputs to a deployed model changes over time.

Technical Definition

When the statistical distribution of inputs to a deployed model changes over time.

How it works

User behavior, market conditions, sensors, and language all evolve. A model trained last year may see inputs today that look subtly different, silently degrading performance. Monitoring input distributions and prediction confidence helps detect drift; periodic retraining counteracts it.

Related Concepts

• Data Pipeline — An automated workflow that ingests, transforms, and delivers data to models.
• Concept Drift — When the relationship between inputs and the target outcome shifts over time, even if inputs look the same.
• Deployment — The process of moving a trained model into a production environment where it serves real users.