Raw data is the source of truth

I keep the original string and import details. If a reading is clearly invalid, I mark it as excluded with a reason. Silently removing the number makes the final statistics impossible to audit.

Sample size matters

A small string can be useful for a function check, but it should not be presented as a strong statistical conclusion. ReloadMastery shows shot count beside every summary so the context stays visible.

Compare like with like: same firearm state, ammunition lot, and environmental context. Combining unrelated strings can create a precise-looking answer that does not describe a real test.

Use statistics as a prompt to investigate

Velocity spread can point to a process worth reviewing. It does not automatically identify the cause. I look at the full record, verify measurements and equipment, and change one controlled variable at a time.