OBJECTIVES:  Current cohort oriented statistical methods are well attuned to analyse efficacy in randomized clinical trials. However, to study real-world patient-level effectiveness in observational settings, more advanced methods should be used. Such methods should allow for product performance based data labelling and identification of confounding factors of real-world performance. Therefore, the aim of this study is to identify such methods, capable of annotating and analysing real-world data. METHODS:  A literature review was conducted to identify statistical methods that enable product performance-based data labelling and that are capable of identifying confounding factors of product performance in a real world context. RESULTS:  Classical survival statistics such as Cox proportional hazard modelling is typically used to estimate patient-level survival gain as drug response. However, Cox proportional hazards models are only useful to model the observed prognostic and predictive markers identified in random clinical trials. Real-world data offers the potential to identify other confounding factors of real-world performance, requiring the use of machine learning techniques. Caution should be taken using real-world data, as measuring product performance in observational studies suffer from selection bias (as opposed to measuring performance in randomized clinical trials). This requires the use of advanced methods like propensity modelling. CONCLUSIONS: Classical survival modelling techniques are useful to identify real-world patient-level product performance whenever they are supplemented with methods for reducing selection bias. To analyse confounding factors of real-world product performance, machine learning techniques will be necessary.