Survival analysis incorporates various statistical methods specific to data on time until an event of interest. While the event is often death, giving rise to the phrase ‘survival analysis’, the event might also be, for example, a reoperation. As such, it is sometimes referred to as ‘time-to-event analysis’. Censoring sets survival analysis apart from other analyses: at the end of the follow-up period, not all subjects have experienced the event of interest, and some subjects may drop out of the study prior to completion. Survival data for a group of subjects is usually visualized by the Kaplan–Meier estimator, representing the probability of a subject remaining free of the event during follow-up. There are several methods to compare survival between the study groups, for example, treatment arms, including the log-rank test and the Cox proportional hazards model. The log-rank test is an unadjusted non-parametric method, whereas the Cox proportional hazards model allows comparison while adjusting for multiple covariates. A principal assumption of the Cox proportional hazards model is that the relative hazard stays constant over time—the so-called proportionality. Specific methods exist for comparison of survival with the general population. This article describes the fundamental concepts every cardiothoracic surgeon should be aware of when analysing survival data and are illustrated with a clinical example.