SQL Server Statistics by Holger Schmeling
A Distribution Statistics object is a sampled map of the distribution of data in a table. Such statistics help the Query Optimizer of SQL Server’s Database Engine to translate the ‘logical’ SQL Query into the best strategy for doing all the physical operations of joining, sorting, and selecting the data. Most of the time, this work happens effectively, but occasionally, if the statistics no longer reflect what is actually in the table, then queries can suddenly run grindingly slow. At this point, the DBA must intervene to correct the problem.