What is AUC?
AUC, short for area under the curve, is a number from 0 to 1 that summarizes a classifier’s ROC curve. It equals the probability that the model scores a randomly chosen positive case above a randomly chosen negative one. An AUC of 0.5 means the model ranks no better than chance, and 1.0 means every positive outranks every negative. Clinical research calls the same measure the c-statistic, and credit risk teams often report the Gini coefficient, which is twice the AUC minus one.
AUC covers every threshold and uses rates calculated within each class, so it depends on neither the cutoff nor the prevalence. That makes it convenient for comparing models before a threshold is chosen. The abbreviation sometimes refers to the area under the precision-recall curve instead, which is more informative when positives are rare, so confirm which curve a vendor means.
AUC in practice
An industrial supplies distributor scores its 4,000 accounts each month for churn risk, and account managers call the top 150. One candidate model has an AUC of 0.81 and another 0.78. The second places more accounts that actually leave in the top 150, because its advantage sits at the head of the list, the only part the team acts on. The distributor chose the second model. AUC is useful for shortlisting models, but the final choice should rest on performance where the business operates. A high AUC also says nothing about calibration, so the scores may be poor estimates of probability.