This advisory condition provides an effective way to determine if statistics are out of date by comparing row count estimates between the control node and the compute nodes for each table in a database.
If the percentage values differ too greatly, APS does not have accurate statistics for that table. This can cause the APS optimizer to choose operations and data movements that are not efficient and result in long query run times.
By default, this condition will evaluate to true if there is at least 20% discrepancy on any table.
When configuring for your environment be sure to import at the target level, not All Targets. Also be sure to update the Database name to the database you wish to evaluate.
Note this condition only evaluates row counts and not cardinality. It also does not take into account whether statistics exist on more than one column, or for which column the statistics exist.
For more info see: https://support.microsoft.com/en-us/help/3046875/how-to-evaluate-pdw-statistics-accuracy