Kinda wish someone would package up capabilities like these as nicely as the SQL Server Profiler and Index Tuning Wizard. But even so, real fixes almost always lie a level above - you can write semantically identical SQL (just like any other language) that has wildly different performance characteristics. That's annoying for a supposedly declarative language. And nothing will save you if you have just messed up your schema - something that tells you do denormalise stuff to get it on the same page of data because of such and such a query load would also help.