Try counting pennies.
If I have 1,754 pennies, I also have an equivalent in dollars, $17.54. I don't actually have dollars, I have the equivalent amount, in pennies, a total count of 1, 754 pennies
A penny is 1/100 of a dollar. So, 1,754 pennies is a total count of 1,754 hundreths (of a dollar).
That value of 1,754 can be communicated as an integer value, a whole number.
When the sender sends a count of 1,754, that raw value is interpreted at the receiver end as 1,754 pennies.
But what if the goal is to know how many dollars the penny value represents?
The receiver could establish a decimal point and infer a dollar sign and produce an equivalent value of $17.54. All done with an integer value
In the industrial comm world, as others have pointed out, some devices allow a decimal point to be fixed for display of an integer value, to do just this. Others don't. Without a displayed decimal point, one is stuck with solutions like adhesive labels on displays, for correct interpretation of the value.
Floating point is best desribed as values that include an exponent (or some call it engineering notation).
1,754 pennies = $17.54
17.54 = 1754 * 10^-2 which is a similar to a floating point number.
In floating point, two values are needed. In the example above, the two values are 1754 (a mantissa) and -2.00 (the exponent)
The sender and receiver would have to agree (in the example) that the mantissa value is assumed to be multiplied by the value 10 raised to the exponent.
Dan