This paper describes a bad field failure data analysis practice where many field failures are ignored or discarded (filtered) before the failure rate calculation is done. This has been called “extreme data filtering.” Several examples are given. A failure rate resulting from this filtering process will be significantly lower than the real numbers. Extreme data filtering is done for different reasons in different situations but always results in very low, very optimistic numbers. These numbers should never be used for Safety Instrumented Function (SIF) verification calculations.