Missing values, dtypes, and the silent model bug

Missing data is not one thing. None, an empty string, a zero, and the word "unknown" all carry different meaning.

The most common beginner mistake is treating every falsey value as missing. That deletes valid zeros. The second mistake is leaving numeric strings unconverted, then wondering why a model or metric behaves strangely.

Good dataframe work writes the policy down: what counts as missing, what gets imputed, what gets rejected, and what type every feature should become.