This is a rule about significant digits. The number of significant digits of the answer is specified by least significant bit in the number involved in the calculation, such as 25.34-24.0. Of these two numbers, 24.0 is accurate to ten digits, so the result can only be accurate to ten digits instead of 1.34.

This reflects the accuracy of the results. It is equivalent to only one tenth of the accuracy when measuring, so the accuracy of the result can't be higher than this, and it is meaningless to write percentiles.

This is mainly used in some experimental sciences.