Estimated vs Actual Rows

Estimated Rows vs Actual Rows

• Estimated Rows: Number of rows SQL Server expects an operator to process. It is calculated using statistics and the cardinality estimator.

• Actual Rows: Number of rows the operator really processed while executing the query.

Why compare them?

If both values are close, SQL Server’s estimate is generally accurate.

A large difference may cause SQL Server to choose:

• An unsuitable join type

• An inefficient access method

• Too much or too little memory

• Unnecessary parallelism

• Sort or hash spills to tempdb

Common reasons for a mismatch

• Missing or outdated statistics

• Parameter sniffing

• Uneven data distribution

• Complex predicates or joins

• Non-SARGable conditions

• Table variables without useful statistics

Example

Estimated Rows: 10
Actual Rows:    500,000

SQL Server planned for only 10 rows, so it might select Nested Loops when a Hash Join could be more suitable.

Key point

A significant difference between estimated and actual rows indicates a cardinality-estimation problem and should be investigated.