# Feature Consistency Check

> Source: https://sukruyusufkaya.com/en/glossary/feature-consistency-check
> Updated: 2026-05-13T20:59:35.636Z
> Type: glossary
> Category: veri-muhendisligi-ve-ai-altyapisi
**TLDR:** A validation process that verifies whether training-side and serving-side feature values are produced with the same logic and definition.

<p>Feature consistency checks are one of the core safeguards against train-serve skew. If the same feature is calculated differently in offline and online systems, model performance can silently degrade. Sample comparisons, hash validation, or definition-level matching checks may be used to detect this. In production AI systems, such checks act as a form of quality assurance.</p>