Abstract: A database management system associates, for one or more database fields, a respective representation of cardinality with different discrete subsets of database records, the subsets preferably being defined by different quantiles of an equal height histogram. The system predicts a relative number of records responsive to a query condition using the representation of cardinality of a quantile in which a query-specified value lies. Preferably, a relative number of responsive records is estimated as a quantile size representation divided by a cardinality representation. The system uses this prediction to determine an optimum query execution strategy. Preferably, the system derives histogram data including cardinality and ordinal numbers corresponding to each quantile using sampling techniques.