1. Confirm there is a problem.
It is often wise to check any calculations, especially those used for purity or mass balance determinations. Check the reproducibility and reliability of the assay method and calculations. Inconsistent sampling can product anomalous results. When monitoring scale-up batches, it may be wise to obtain another sample from the reaction; then prepare another sample for IPC analysis and reassay. (from Practical Process Research and Development)I can't tell you how many times that I've begun to feel that rising feeling of panic, only to get another sample or redo an assay and get a great big sigh of relief.
(Of course, if your second sample confirms your bad result, don't go check a 3rd time, or a 4th, or a 5th. Sometimes bad news is really bad news.)