Accuracy may be defined as being right, or having a valid measure. That means, while there might be a cone of uncertainty about whether the instrument is recording what we think it is measuring, the instrument is reflecting a valid version of reality.
Precision may be defined as being very granular, without regard to validity. That means the instrument exchanges volatility in the reading for accuracy in the reading. A precise instrument is consistently wrong.
Some would prefer a thermometer that is precise to within two tenths of a degree, regardless of how accurate it is. Others would prefer a thermometer that rounds to the nearest 5 degrees (plus or minus 2.5 degrees, 2 sd).
I prefer accuracy. If I’m trying to decide which coat to wear (Ie. Make a decision), being within 2.5 degrees, 2sd is adequate than believing its 4.55 C but having it feel like -22.
In other words, to me, it’s far more important to be right, to have a valid result from which to make decisions, than it is to be precise – to have something that is likely to produce a reliable, consistent, result. The reason why web analysts get into arguments with marketing scientists on the topic is because it goes to a deep value root.
