Virtually any use of FLOAT, REAL, or DOUBLE PRECISION data types is suspect. Most applications that use floating-point numbers don't require the range of values supported by IEEE 754 formats. The cumulative impact of inexact floating-point numbers is severe when calculating aggregates. Instead of FLOAT or its siblings, use the NUMERIC or DECIMAL SQL data types for fixed-precision fractional numbers. These data types store numeric values exactly, up to the precision you specify in the column definition. Do not use FLOAT if you can avoid it.