SQL Server Execution Plans - Red Gate Software

Chapter 1: Execution Plan Basics
Why the actual and estimated execution plans
might differ
Generally, the estimated and actual execution plans will be the same, in terms of the
operations used. However, circumstances can arise that can cause differences between
the estimated and actual execution plans.
When statistics are stale
The main cause of a difference between the estimated and actual plans is differences
between the statistics and the actual data. This generally occurs over time, as data is
added and deleted. This causes the key values that define the index to change, or their
distribution (how many of what type) to change. The automatic update of statistics that
occurs, assuming it's on, only samples a subset of the data in order to reduce the cost
of the operation. This means that, over time, the statistics can become a less and less
accurate reflection of the actual data. Not only can this cause differences between the
plans, but you can get bad execution plans because the statistical data is not up to date. 5
When the estimated plan is invalid
In some instances, the estimated plan won't work at all. For example, try generating an
estimated plan for the simple bit of code in Listing 1.20, and you will get the error shown
in Listing 1.21.
The optimizer, which is what generates estimated execution plans, doesn't execute
T-SQL. It does run the statements through the algebrizer, the process outlined earlier that
is responsible for verifying the names of database objects.
5 An example demonstrating how a drastic change in the data can affect the execution plan is given in the Statistics and indexes
section of Chapter 4.
59