The Baseline You Must Beat

Linear regression fits the best straight line through continuous targets — predict revenue, demand, dosage. Logistic regression runs the same linear score through a sigmoid and outputs a probability — predict churn, default, click.

Despite the name, logistic regression is a classifier. Both are convex — there's exactly one best answer, and training finds it in milliseconds. And every coefficient reads as a sentence: 'each extra year of tenure adds this much to retention odds.'