When you design a program, there is an important step consisting of choosing the right type for each variable.
Sometimes, the choice is constrained by external factors. This happens when, for instance, you use the Arduino with an external sensor able to send data coded as integers in 10 bits (210 = 1024 steps of resolution). Would you choose byte type knowing it only provides a way to store number from 0 to 255? Probably not! You'll choose int.
Sometimes you have to choose it yourself. Imagine you have data coming to the board from a Max 6 framework patch on the computer via your serial connection (using USB). Because it is the most convenient, since you designed it like that, the patch pops out float numbers encapsulated into string messages to the board. After having parsed, cut those messages into pieces to extract the information you need (which is the float part), would you choose to store it into int?
That one is a bit more difficult to answer...