I thought about that as well, having the Clustered index be Date,PkID (but that still would be as wide as my other key which was date,int,int), my PK is bigint...

Yea, if i put those in the NC index, then the number of logical reads increases and it won't use the NC index.

so PK ID - clustered

date,aid,bid or aid,bid,date (NC)...

Gail,

The nonclustered index is currently a PK ID, as for the table, there are many columns, but it's not OVERLY wide (there's one column that I've already suggested we need...

I'm pretty sure I need to have datetime in "some" index, I agree with GilaMonster as it's the main part of the query to retrieve a set of rows. ...