Categorical or numerical?

Data comes in two broad kinds. Numerical data measures a quantity, a price, a weight, a count, and arithmetic on it means something. Categorical data labels a group, a region, a department, a product code, and arithmetic on it means nothing at all.

The trap is that a lot of categorical data is written in digits. A postcode, an order number, a phone number and a year group are all labels that happen to look like numbers. Excel cannot tell the difference and will happily average your postcodes for you.

The test is simple: would the halfway point between two of these values mean anything? Halfway between £10 and £20 is £15, which is a real price. Halfway between postcode 2000 and postcode 3000 is not a place.