You work for a magazine publisher and have been tasked with predicting whether customers will cancel their annual subscription. In your exploratory data analysis, you find that 90% of individuals rene

In an imbalanced dataset, simply having higher accuracy for the majority class (renewers) is easier to achieve and less indicative of a good model than high accuracy for the minority class.

Predicting everyone renews would give 90% accuracy but would completely miss all cancelations; a model that correctly identifies 99% of the difficult-to-predict cancelers is performing significantly better than a naive majority-class predictor.

Predicting the minority class (customers who cancel, only 10% of the data) with 99% accuracy is generally considered a very good result, as this class is often much harder to identify correctly compared to the majority class. The overall utility of the model is often focused on identifying these rare, critical events, making high precision/recall for the minority class highly valuable.

While 80%+ accuracy across both groups seems good at first glance, this statement oversimplifies the problem; in imbalanced classification, evaluating minority class performance is critical, and 99% on the harder class is a strong indicator of value.