Relative percentage difference (RPD) is a way of measuring the variation in a set of data that looks at the variation as a proportion of the average or target value. For example, it may be used for quality or portion control in a factory that makes 1 kilogram (kg) and 10 kg bags of widgets. A 1 kg bag that was 100 g too light would have a greater RPD than a 10 kg bag that was 100 g too light; 10 and 1 percent respectively.
Determine what you are calculating RPD against. In the factory example, you would calculate the RPD against the stated weights of the widget bags, 1 kg or 10 kg. If you don't have a target, the most useful thing to calculate the RPD against would be the average value.
The RPD for a particular item in your data set is calculated like this: RPD = ((x - t)/t) * 100. Here, x is the value for the item -- say, your bag of widgets -- and t is the "target" value, either the actual target if there is one or just the average value. For example, for a bag of widgets that should weigh 1 kg, but in fact weighs 950 g: RPD = ((950g - 1000g)/1000g) * 100 = -5 percent.
From the resulting RPDs -- you can calculate an RPD for each item in the data set -- you can understand how large the variation is. In particular, if most of the widget bags have a very small RPD but a few have large RPDs, you would probably conclude the factory process is generally accurate, but sometimes significant mistakes are made.
RPD can be positive or negative, depending on the value relative to the target, so if you get a negative value, this does not mean you have done it wrong.
RPD is used in a similar way to relative standard deviation (RSD).
It is relatively easy to use a spreadsheet to automate the calculation of RPDs.