Implementation of the sparse column makes sense only if the data in the column is, well, sparse. This is because it takes more space to store not null values in the sparse columns. For example, character types, such as char and varchar need twice as many bytes to store not null data if the column is defined as sparse. For those types only if 60% or more of all records are nulls then it makes sense to define the respective columns as sparse. For example, many databases storing consumer/customer/person records might end up having a column storing data for so-called address line 2 which will not have any data in the majority of records, which makes such column a good candidate to be defined as sparse.

Oleg