# Overfitting · Glossary
When a model learns the noise and specific quirks of its training data rather than patterns that generalise, so it performs well in development and poorly on new data.
[Glossary](/glossary) · Machine learning and MLOps

# Overfitting

     When a model learns the noise and specific quirks of its training data rather than patterns that generalise, so it performs well in development and poorly on new data. It is one of the most common reasons models disappoint after deployment.

## Why it matters

     An overfitted model can produce impressive test results that do not survive contact with real operations. Leaders relying on development metrics may approve a model that is less reliable than it appears.

     Overfitting is prevented by honest evaluation: testing on data from later time periods or different sites, keeping evaluation data separate from all development decisions, and preferring simpler models when performance is similar.

## In practice

     For example, a UK construction contractor’s cost-overrun model might score highly when tested on a random sample of past projects but poorly on the most recent year. Testing on later projects reveals that the model had memorised details of a few large jobs.

## Where Rodan fits

     Rodan validates models on data that reflects how they will be used, as part of [Applied AI Engineering](https://rodan.io/applied-ai-engineering).

## Related terms

- [Machine learning (ML)](/glossary#machine-learning)

- [Supervised learning](/glossary/supervised-learning)

- [Feature engineering](/glossary/feature-engineering)

- [AI evaluation (evals)](/glossary#ai-evaluation)

- [XGBoost](/glossary#xgboost)
HTML: https://rodan.io/glossary/overfitting