Clinical endpoints, such as overall survival, directly measure relevant outcomes. Surrogate endpoints, in contrast, are intermediate, stand-in measures of various tumour-related metrics and include tumour growth, tumour shrinkage, blood results, etc. Surrogates may be a time point measurement, that is, tumour shrinkage at some point (eg, response rate) or biomarker-assessed disease status, measured at given time points (eg, circulating tumour DNA, ctDNA). They can also be measured over time, as with progression-free survival, which is the time until a patient presents with either disease progression or death. Surrogates are increasingly used in trials supporting the marketing authorisation of novel oncology drugs. Yet, the trial-level correlation between surrogates and clinical endpoints—meaning to which extent an improvement in the surrogate predicts an improvement in the direct endpoint—is often moderate to low. Here, we provide a comprehensive classification of surrogate endpoints: time point measurements and time-to-event endpoints in solid and haematological malignancies. Also, we discuss an overlooked aspect of the use of surrogates: the limitations of surrogates outside trial settings, at the bedside. Surrogates can result in the inappropriate stopping or switching of therapy. Surrogates can be used to usher in new strategies (eg, ctDNA in adjuvant treatment of colon cancer), which may erode patient outcomes. In liquid malignancies, surrogates can mislead us to use novel drugs and replace proven standards of care with costly medications. Surrogates can lead one to intensify treatment without clear improvement and possibly worsening quality of life. Clinicians should be aware of the role of surrogates in the development and regulation of drugs and how their use can carry real-world, bedside implications.