Overfitting.

Overfitting occurs when a model learns the training data so thoroughly—including its noise and random variation—that it fails to generalise to new examples. An overfitted model achieves near-perfect accuracy on training data but performs poorly on validation or test data, because it has memorised specifics rather than learned general patterns.

Detecting and Preventing Overfitting

Overfitting is detected by comparing performance on training vs. validation data—a large gap indicates the problem. Prevention techniques include regularisation (penalising model complexity), dropout (randomly disabling neurons during training), early stopping (halting training before performance degrades), and cross-validation.

Analogy for Marketers

Overfitting is analogous to a marketing strategy optimised so heavily for past campaign data that it stops working in new conditions. A headline A/B test that picks the winner based on too-small a sample might 'overfit' to random variation rather than true user preference—future campaigns underperform because the chosen variant wasn't genuinely better.