Precision & Accuracy

This came across this on the IXP list today from Don Wells and I thought it was one of the better descriptions of precision and accuracy.

One thing we all need to remember when talking about metrics is the difference between accuracy and precision.

Consider a yardstick that I take and cut a foot off of.  I then repaint it and put the markings back on.  So I have a yardstick that is only 2 feet long instead of 3 feet.  This yardstick is not accurate but it is precise. That is, I can measure something twice and get the same measurement.  I can
measure two things add the measurements together and expect the sum to equal a single measurement of the two things stuck side by side.  I can add these measurements because they have precision, unfortunately all the measurements are wrong because they are not accurate.

Now let's take a thick piece of elastic slightly less than 3 feet long and mark it like a yardstick.  To measure we stretch it out just a little bit and measure.  This yardstick has accuracy but not precision.  That is, when I measure something it is real close to the correct number but every time I
measure it is slightly different.  It has low precision.  If I add measurements together the precision gets even worse.

Most of the measurements we make as software people have accuracy but not precision just like the elastic yardstick only worse.  This is because we deal with people and subjective numbers.  The point of all this is that many of these numbers don't work well in complex formulas.  That is, you can not add, subtract, multiply or divide these numbers.  It is often better to find a first order measurement with good accuracy and reasonable precision instead of doing lots of math on numbers with little precision.