This all is useful for the query optimizer (QO going forward) to determine and craft the proper seeks, scans, lookups, joins of various kinds, etc., along with the amount of memory to allocate for the processing of results. Statistics provide insight into what values exist in a domain of data, how discrete (aka different, unique) the values are from one another, and what their storage size and counts are. And those plans are inappropriate because they're based on a lack of understanding data values. That metadata is used to create and compile the best execution plans for the queries being submitted by end users against your databases. If you have performance problems with your SQL Server instances, many of them can be linked to the query optimizer making bad decisions when building execution plans.

In part one of this series, I explained that statistics provide your data values' metadata. Now it’s time to see what those automatic operations leave in their wake as well as looking at when it makes sense to manually add statistics objects to your tables & indexes in Microsoft SQL Server.

In the first article in this series on SQL Server distribution statistics, I covered at a high level what statistics are, why they are important, and the options at the database level for creating and updating statistics automatically.

Update statistics sql server