"Jesper Jensen" <moellemand@post.tdcadsl.dk> wrote in message news:<40732877$0$274$edfadb0f@dread12.news.tele.dk >...
> Thanks, Simon, that one did the trick.
>
> One less mystery.
>
> On my machine, QA tells me that the two queries (the index scan on
> ProductID/Quantity vs. the clustered index scan) takes 43.65 and 56.35% cost
> respectively. I would argue, that this saving is not worth the 'used up
> index space'. In my professional life, I've seen tables, which are wide
> enough (200+ columns) to demand, that precious index space is saved.
>
> Basically, I think there are too many 'ifs' before an auto-created index
> saves performance, but I appreciate the optimization idea behind it.
>
> Thanks - Jesper
>

<snip>

Well, you have to be careful about reaching conclusions based on
simple queries using small data sets. It's possible that a complex
join involving millions of rows would give a more significant
difference. To get a definite answer for your environment, you would
have to do some benchmarking, with and without statistics.

Simon