Beware of using percentages!

Percentage (%) is one of the popular measures used in the data world. While it is a very good indicator of the distribution of data points for a single variable/ dataset, it is a confusing measure when used with multiple variables/ datasets when combining them into a unified source of truth. Let us consider two examples, one for each case to understand it better.

Case 1: % distribution of people liking different ice-cream flavors from strawberry, vanilla, pista, and chocolate out of 1000 people surveyed for the same.

Case 2: % distribution of people liking different ice-cream flavors from strawberry, vanilla, pista, and chocolate out of 1000 people surveyed for the same at two different locations.

As we can see from the above tables for Case 1 and Case 2, Case 1 gives a clear picture of the overall distribution of interests of people for the variety of ice-cream flavors. In case 2, a combined value of percentage can only be a derived value from absolute values! So, the actual table would look like the below one when the data from two locations is combined: