In some cases, they help SQL Server generate better execution plans based on actual row counts. The best practice isn't "always create the clustered index first" or "always create it later." The real ...