Dimensionality Reduction abounds in ML pipelines. While we can expect more information is lost when dimension is reduced more, the phenomenon remains a conceptual mystery in theory. In this work, we rigorously quantify such empirical observations using geometric techniques.