MS SQL Consulting

SQL Query Plan Parser

XML
INPUT — Paste XML query plan here
OUTPUT — Analysis
GRAPH — Plan Visualizer
←
Paste & Parse
Paste SQL Server execution plan XML on the left and click Parse to extract insights.

Free SQL Server execution plan analyzer

Paste the XML of a SQL Server execution plan (.sqlplan) and the parser explains what is actually slow and why. Everything runs locally in your browser — the plan is never uploaded.

What the parser checks

How to get the execution plan XML

  1. Actual plan in SSMS: enable Include Actual Execution Plan (Ctrl+M), run the query, right-click the plan and choose Show Execution Plan XML — or save it as a .sqlplan file and paste its contents.
  2. From the plan cache: SELECT qp.query_plan FROM sys.dm_exec_cached_plans cp CROSS APPLY sys.dm_exec_query_plan(cp.plan_handle) qp;
  3. From Query Store: the query_plan column of sys.query_store_plan.

Plans from the cache and Query Store are estimated plans. Actual row counts, reads, spills and waits are only available in an actual plan.

FAQ

Is my execution plan sent to a server?

No. The XML is parsed with JavaScript in your browser. Nothing is uploaded or stored, so plans containing table names or parameter values stay on your machine.

Why are estimated and actual rows so different?

Common causes are outdated or missing statistics, parameter sniffing (the plan was compiled for a different parameter value), table variables, local variables, functions in predicates and correlated columns. The parser highlights the operator where the estimate first goes wrong — that is usually where to start.

What is an implicit conversion and why does it matter?

When a column and the value it is compared to have different data types (for example a varchar column compared to an nvarchar parameter), SQL Server converts one side. If the column side is converted, an index seek may become a scan. Fix it by matching the parameter or literal data type to the column.

What does a spill to TempDB mean?

A Sort or Hash operator did not get enough memory and wrote data to TempDB, which is much slower than memory. The usual root cause is a row underestimate below that operator, so the memory grant was sized too small.

Should I create every missing index the plan suggests?

No. Missing index hints are generated per query, ignore existing similar indexes and do not consider write overhead. Combine overlapping suggestions, check the column order and compare with the existing indexes before creating anything.

Built by Aksel Rohtsalu, MS SQL Server consultant at dba.ee. Related tools: Statistics IO/TIME Parser · SQL Server Health Check.