So, we left off with swatching. When my swatches aren't lie-y lying liars, they are extremely helpful. Heck, even when they do lie, they're helpful. I know most of the time I'd just like to dive in on knitting the PROJECT, but we all know the heartbreak that can ensue. Plus, when I'm designing, the swatch is the cornerstone of the design.
After the swatch comes the math. Luckily for me, it's basic math, which my parents would quickly point out has always been one of my strengths. Once we got to imaginary numbers, I quit, but I'm seriously quick on the basics. Usually, I calculate the cast-on number for my key size, whether that's me or the size I've contracted to produce for the sample. For now, most of my designs are bottom-up, so I start with the hem. Since, in addition to working from the hem, I prefer working in the round, I figure out the start of the sleeve right away and usually cast on a sleeve first. That way I finish one sleeve right away, while I'm "full of prance" as Elizabeth Zimmermann said. Plus, I won't end up on sleeve island or suffer from second sleeve syndrome (cousin to the dreaded second sock syndrome or SSS).
I know some designers who write up the entire pattern before they knit a stitch (aside from the swatch), but I prefer starting with some calculations, then take notes as I go. Sometimes the design happens on the needles. Oftentimes I'm still working out an idea (usually a crazy, I should be able to do this with knitted fabric kind of idea, like adding ribbing for shaping in Turn of the Glass). As a designer I'm most interested in manipulating the knitted fabric to do what I want it to do, or what I think it should be able to do. And exploring the brilliant engineering that EZ pioneered (if you're interested, you should check out her Knitting Workshop). An Aran for Frederick came about because I was curious to see how cables would work into her hybrid yoke idea.
Once I've successfully finished the piece, I consult all the notes I've taken, mostly in my Moleskine notebooks, and write up the pattern for the size I've knit, leaving spots for the other sizes. And then it's time for the spreadsheet! I'm no Excel power user, but I manage to get my OpenOffice spreadsheets to calculate everything for the various sizes. Usually, I have rough numbers to start and then I massage them to make things smooth for all you knitters. For instance, I will have to adjust the number of stitches between cables, but not too much or it will start to look like a different sweater. I started out looking at Marnie's amazing spreadsheets (that woman really knows how to make Excel do her bidding!), though I ended up creating my own since hers was for a sweater knit in pieces.
When I'm massaging those numbers, I refer to the brilliant chart Ysolda compiled for different sizes [link], which has been extremely helpful. Grading patterns is a challenge. You don't just add an inch or two all around in all directions for each size. You do have to finesse it. I've also built a lot of checks into my spreadsheets, converting stitch counts back into measurements to make sure we haven't veered into crazytown.
And the really cool thing I've figured out sort of how to do is calculate yardage requirements. I was really worried about doing this when I started designing sweaters. It seemed like it would just be guesswork. It is a somewhat inexact science, since knitters take up yarn into stitches at different rates, but still. I keep careful track of how much yarn I use in the sample, then use the spreadsheet to come up with some calculations as to how many stitches went into the whole piece. I always make these a little generous, so you have a cushion. Plus, I wouldn't want to calculate each and every stitch, especially with shaping in some garments. Anyway, I was very excited when I figured out that part of designing and grading.
Happily, all this crazy spreadsheet work has, thus far, meant that my patterns are pretty clean as far as my numbers go. Of course, after it's all written up, it goes off to my tech editor to make sure things are correct.