Largely inherited from the past and from strictly internal laboratory use, the certification of reference material using round-robin assaying is often insufficient for the use made today for external assay laboratory control by client parties. Certification often needs to be revisited on the basis of existing round-robin data, especially to derive reproducibility parameters to the end of establishing probabilistic performance gates for operational error detection on a batch per batch basis (ie batch-rejection'). The most common difficulty arises when only a few certification samples have been analysed in one single batch by each laboratory, so that the round-robin data do not contain the full variability that needs to be quantified. Existing certification data sets from properly submitted round-robin samples were studied using analysis of variance to derive a remedial method, leading to the proposal of default correction factors to restore full variability estimates when no other avenue is available.