PCR optimization

Anyone here do a lot of PCRs (Polymerase Chain Reactions)?  I was wondering what steps you take to optimize your PCR, what kind of gradients you do, and how many PCRs it generally takes you to get to your final optimized conditions.

If it matters, you're PCRing up a plasmid, and you're also PCRing an insert out of a bacterial genome.

I guess I'll say what I do...

First I start off with an annealing temp vs [Mg2+]

After finding the best anneal temp, I do a triple gradient: [Mg2+] vs [DMSO] vs [BSA]

After getting the best concentration of the adjuvants, I then do a [Mg2+] vs [Buffer] gradient.

I'm wondering about the triple gradient... should I be doing that after or before I find the best buffer concentration.

Also, I never change [dNTP] or [primer]; I keep them at (200uM) and (300nM) respectively.

Are these all necessary variables to tinker with, or could I simplify my optimization without losing too much in terms of yield?