Why Good Research Measures Its Uncertainty

A number without an interval is a rumour with a decimal point. It tells you what was computed and nothing about whether it would survive being computed again — on a different sample, with a different instrument, by someone else.

Reporting uncertainty is often treated as an act of modesty. It is not. It is the part of the result that carries the most information, because it tells you what the result can be used for.

1. The world. Genuine variability in the system — the thing you are usually trying to characterise.
2. The instrument. Noise, drift, resolution limits, and everything the instrument silently fails to see.
3. The model. Error introduced by the simplification you chose, which does not shrink when you collect more data.

Pooling these into a single error bar is common and costly, because they have different remedies. More data reduces the first. A better instrument reduces the second. Only a different model reduces the third — and no amount of sampling will tell you that is what you needed.

This is also the honest answer to why an early-stage research company should publish so little so slowly. Producing a number is fast. Producing a number you can defend the interval on is not, and only the second one is worth anyone's time.