So far, we have had to analyze query plans by correlating information in plan and operator properties to create working hypotheses on how to solve query performance issues. One constant throughout all these troubleshooting scenarios has to do with comparing estimated rows with actual rows flowing through the operators in a query plan. This is because significant differences between estimated and actual rows usually expose Cardinality Estimator (CE) issues, which indicate several possible causes, from outdated statistics, to parameter sniffing or even out-of-model constructs such as User-Defined Functions (UDF) or Multi-Statement Table-Valued Functions (MSTVF).
And depending on the query performance problem, it may not ...
