ML models are only as good as the data they're trained on. Data quality is the single biggest factor in model performance.

What Makes Good Training Data

• Volume: Enough examples to learn patterns
• Quality: Accurate, correctly labeled, minimal noise
• Diversity: Representative of the full range of real-world inputs
• Balance: Appropriate distribution across classes

Common Data Problems

• Class imbalance — 99% negative, 1% positive examples
• Label noise — mislabeled training examples
• Data leakage — future information in training data
• Distribution shift — training data doesn't match real-world data

Reference:

Google ML data preparation guide

https://developers.google.com/machine-learning/data-prep