I will make a confession: I am yet to write a book series by planning and design (that is, drawing a story arc that initially spans more than one book).
With Quest of the Messenger, I started all backwards - from what eventually became the second book of the trilogy - proceeded to book 1 because the mythology and back story were just too good to pass, and eventually finished with book 3, because it was quite impossible to wrap up the whole thing without it.
In Wild Children, I encountered even more of a challenge, because I had really set my heart on a standalone novel, and found it hard to accept the truth when a wise friend pointed out that I'm simply trying to cram too much into one book (albeit one of close to 120K words). So I began working on the sequel - even before the first book was published - and, as soon as it was done, realized at least another book is needed to make the story complete.
I know several reverse cases - book series that really could, and should, have been wrapped up in a single volume, or series that started off with a good pace, were successful, and then began to drag because the author and publishers simply couldn't relinquish the steady income of an audience waiting for the books. It is eventually a big letdown, though, and unfair to the readers, who feel like they are being duped, and justly resent it.
Ideally, I believe it's probably best to plan a series before starting to write it - at least to the extent of how many books it will include, and the approximate outline of each volume. I know that, had I initially planned Wild Children as a series, I would have chosen to do a few things differently in the first book, but as it was already written when I realized there would be a sequel, I had to work around this.
I guess it all comes down to a simple but tricky principle: know when to start, and know when to stop. Don't be afraid to jump ahead into a sequel, but don't let a series drag on when it's obviously done all it could, either. Also check out this great post about writing series.