Judgments about evidence and the resulting recommendations in healthcare are complex, especially in rapidly evolving fields such as reproductive medicine. Case reports have transformed the treatment of infertility from the first reported case of successful IVF, through ICSI and PGD to uterine transplantation, and in many of these cases there is no need to conduct a prospective trial. Indeed, in many cases it would be impossible or unethical (1). However, there are many interventions where the benefit, if any, is less clear. Recent examples in the treatment of infertility are the use of endometrial scratching or the use of PGT-A in IVF where initial enthusiasm has ultimately been replaced with disappointment. The reason for this change is the inherent bias (both known and unknown) in lower evidence level studies – such as cohort studies or ‘big data’ studies where the different interventions compared are often applied to groups with different prognostic profiles.