Data modelers face a choice when encountering multiple variations of a data item. The item itself is unimportant; it could be a name, a physical address, a phone number, an email address, a shoe size, ...