Database Administrators Stack Exchange is a question and answer site for database professionals who wish to improve their database skills and learn from others in the community. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I have the following query

Select  Pt.PRODID, PT.INVENTREFID,  Inv.ItemName, PT.ItemID, Configid,  Pt.QTYSCHED,
     WHEN 'WJB' THEN 'Sub - Assembly'
     ELSE 'Stock'
  END as CustomerName
from Dynamicsv5Realtime.dbo.PRODTABLE PT
Join Dynamicsv5Realtime.dbo.Inventdim ID
On PT.InventdimId = ID.InventdimID
and PT.Dataareaid = ID.dataareaid
Join Dynamicsv5Realtime.dbo.INVENTTABLE Inv
On Inv.itemid = PT.ItemId
Left Join Dynamicsv5Realtime.dbo.SALESTABLE ST
and ST.Dataareaid = PT.dataareaid
where pt.PRODSTATUS in(2,3,4)
Order by 7 

when I was looking to improve it I came up with the following index


However this was not as efficient as the following which the missing index DMV came up with


What I am wondering is when is it more efficient to put columns in the index itself rather than in the Include part of the index?

May thanks

share|improve this question
up vote 9 down vote accepted

Very generally and non-technically, you should consider using the column as part of the index key if you are going to use that column in the WHERE clause (or otherwise filter/seek), and INCLUDE if it is merely being used to avoid a lookup (I like to call this "coming along for the ride"). Of course there are always exceptions:

  • you can't always include columns in the key due to size;
  • it may not matter due to cardinality; or,
  • it might not affect the plan due to where you would have to put it in the key to allow the index to continue satisfying other query semantics.

There are also cases where an INCLUDEd column can satisfy a filter just fine, and plenty of cases where changing the index might help this specific query but it might wreak havoc on the rest of your workload. You should always test index changes against an entire workload representing as much of a full business cycle as possible, rather than rely on the missing indexes DMV or your own analysis which, it seems, is focused on a single query.

I'm sure Paul White will come along with a much more elaborate and technically accurate response.

share|improve this answer
Thanks for that Aaron, do you have any articles on how to go about understanding what a full workload is and how to capture one? – Tom Mar 8 '13 at 15:29
@Tom you can use a server-side trace to capture a workload. You'll probably read a whole lot about Profiler; don't use it against production servers, PLEASE. – Aaron Bertrand Mar 8 '13 at 15:58
@Tom Depending on what version of SQL Server you are using you can also leverage Extended Events in lieu of SQL Trace. – Thomas Stringer Mar 8 '13 at 17:49
@Thomas while true, I don't think you can feed Extended Events data to tools like the database engine tuning advisor or distributed replay. – Aaron Bertrand Mar 8 '13 at 17:56
@AaronBertrand Good point, I didn't read the contextual usage. – Thomas Stringer Mar 8 '13 at 18:17

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.