A computer stores everything as patterns of bits, so it needs to know how to read each pattern. The data type is that instruction. The same symbol 5 can be the number five, which you can add and multiply, or the character '5', which is just a piece of text like 'A'. The quotation marks in code tell you which: 5 is an integer, '5' is a string.
The type decides what each operation means. For integers, + adds: 5 + 3 = 8. For strings, + joins: 'ice' + 'cream' = 'icecream', and '5' + '3' = '53'. For booleans, the operations are AND, OR, and NOT. Some mixtures are refused: in Python, '5' + 3 is an error, because the language will not guess whether you meant to add or to join. Division can surprise you too: in Java or C, 7 / 2 with two integers gives 3, dropping the remainder, while 7.0 / 2 gives 3.5.
A value that looks like a number is not always one. A phone number such as 0412 555 019 is stored as a string, because nobody adds phone numbers and a number type would drop the leading 0. Typed input usually arrives as a string too, so age = input() followed by age + 1 fails or misbehaves until you convert it with int(age). Ask what a value is for, not just what it looks like.
Think of it like this. Data types are like the units on a measuring jug and a ruler. Both show numbers, but adding 3 cups to 3 centimetres makes no sense, so you check what each number measures before combining them. A computer checks types the same way. The comparison stops at strings: text measures nothing, and joining two pieces of text just sets one after the other.
Core idea. Operations depend on type: you can add numbers, concatenate strings.