The approach
Statvis ran its PCB congener workflow and visualizations across the full dataset. A small cluster of samples stood apart from the rest of the site's data, even though those samples had already passed conventional data validation.
The investigation
The cluster traced back to a single analytical batch. Its samples came from locations like those of other samples and were analyzed within days of neighbouring batches, so an environmental source couldn't explain a distinct signature. The batches analyzed just before and after showed consistent congener patterns; this batch's profile didn't make chemical sense for the site.
The result
The affected results were dealt with before they could influence any conclusions, and the corrected dataset went on to support three years of site investigation.
Validation checks results one at a time. Comparing whole congener profiles across a dataset catches what only shows up in the pattern.