This is a story of a 0.5% discrepancy. When is it good enough?
In the 1890s, a 0.5% discrepancy in the density of nitrogen caught the attention of chemist William Ramsay and John Strutt (Lord Rayleigh). At the time, air was known to be nitrogen and oxygen, along with a few other components such as carbon dioxide and water vapor.
But nitrogen isolated from air had a consistently different density compared to nitrogen produced chemically. Many would have dismissed it as experimental error. Ramsay and Rayleigh didn't.
That tenacity led them to discover argon, and eventually an entire family of elements: the noble gases. It earned them a Nobel Prize, though not without pushback. Mendeleev resisted because they didn't fit his periodic table. Confirmation bias in action, I suppose.
One of those noble gases is radon, a colorless, odorless, radioactive gas and the second-leading cause of lung cancer. The same scientific curiosity that uncovered a 0.5% anomaly inadvertently gave us the knowledge to detect something that kills thousands every year.
So, is 0.5% good enough? It depends on the context and the consequences of what you're missing.
In pipeline integrity, we face this question constantly. Sometimes a 0.1% signal in ILI data might be dismissed as noise, or it might be the early warning of geohazard-driven movement that leads to a rupture.
Ramsay's story reminds us: epistemic uncertainty, which is not knowing what could have been known, can have mortal consequences.
Check your radon.