sampling percent

sampling percent is the numeric value of the sampling percentage, such as 5 for 5%, 10 for 10%, and so on.

Summary Information

Default value


Range of values

0 – 100



Display level


Required role

System or database administrator

Configuration group

Query Tuning

To reduce I/O contention and resources, run update statistics using a sampling method, which can reduce the I/O and time when your maintenance window is small and the data set is large. If you are updating a large data set or table that is in constant use, being truncated and repopulated, you may want to perform a statistical sampling to reduce the time and the size of the I/O.

Remember that with sampling, results are not fully accurate. Balance changes to histogram values against the savings in I/O.

Although a sampling of the data set may not be completely accurate, usually the histograms and density values are reasonable within an acceptable range.

When you are deciding whether or not to use sampling, consider the size of the data set, the time constraints you are working with, and if the histogram produced is as accurate as needed.

The percentage to use when sampling depends on your needs. Test various percentages until you receive a result that reflects the most accurate information on a particular data set.

Statistics are stored in the system tables systabstats and sysstatistics.