What is 'loss of information' in classified data ?

The classification of data as a frequency distribution has an inherent shortcoming. While it summarises the raw data making it concise and comprehensible, it does not show the details that are found in raw data. There is a loss of information in classifying raw data though much the latter provides only the number of records in that class (i.e. frequency = 6) but not there actual values. All values in this class are assumed to be equal to the middle value of teh class interval or class mark (i.e. 25). Further statistical calculations are based only on the values of class mark and not on the values of teh observations in that class. This is true for other classes as well. Thus the use of class mark instead of the actual values of the obervations in statistical methods involves considerable loss of information.