📏An interval is more honest than a point

A single number claims more than I know. Suppose I say that triage time fell by 18 percent. It sounds like a measurement. What I have is an estimate from a sample, and the truthful version is something like between 6 and 30 percent, most likely near 18. The point is the middle of the range with the range cut off. Reporting only the point is a small lie of omission, and it is the listener who pays for it, by deciding as though I were certain.

So the order matters: the range first, then the middle. If I say eighteen and add the caveat, the eighteen is already in the room and the caveat is heard as modesty. If I say six to thirty, people do the right thing unprompted. They ask whether six would still be worth it.

That question is the whole use of an interval. It turns is it good into would the decision change anywhere inside this range. If not, decide and move on; the uncertainty is real and irrelevant. If so, the width tells me what more data would buy.

I first did this properly on 8 May, presenting the triage test as an interval where the template wanted a win. The discussion that followed was about the decision and not about me.

There is a kindness in it too. A range tells the next person how hard they may lean on my number. A carrier scored on six loads and one scored on six hundred should not appear as two equally confident ranks.

Where it fails: an interval covers sampling error only. It says nothing about a biased sample or the wrong measure, and a tidy range can make a bad number look more trustworthy than it is.