We delve into the statistical inference limitations, especially with respect to experimental results derivations, and how causal inference provides a formal approach that generalizes randomized controlled trials for the scientific inquiry. We also develop an interface to bridge between the well known frameworks of Ananke and Pomegranate, to perform asymptotically debiased effect estimations, from surrogate experiments, in absence of hidden confounders. Thus, we analyze how relevant errors of finite sample size are, to see if a pilot study may effectively be used to decide the best. Lastly we examine other causal tasks and limitations of current approaches and the one we propose, as well as possible future improvements.