Access to blinded interim data has provided opportunities for performing sample size re-estimation (SSR) in a blind fashion. In this approach, while the assumed treatment effect at design stage is held unchanged, knowledge about nuisance parameters relevant to sample size calculation are updated using interim aggregated data and the sample size is re-calculated thereafter. Blinded SSR designs are known to have little regulatory risk and pay no penalty in final data analysis. However, if the target treatment effect is of great uncertainty or the true effect is less than what is clinically meaningful, blinded SSR could either be helpless or result in an erroneous action of sample size increase. This presentation intends to clarify situations when a blinded SSR design is suitable and useful, and when an unblinded SSR, although more complicated statistically and operationally, is the way to get sample size just right. It also highlights principles in handling noise associated with re-estimated parameters and preventing miss-step of sample size increase simply based on re-estimation of nuisance parameter(s).