A switching layout:
1.a. Is it a mainline switching (mainline wayfreight)?
b. Is it a branchline switching (no road freights or passenger schedules)
2.a. Is this going to be a scenicked layout?
b. Is this going to be a "switching problem layout"
( will scenery aid, or interfere with the switching? )
c. will you use manual, or "automatic" uncoupling ?
3. (you indicated "city", so tight curves, and not a lot of open space) What kinds of industries ? What unique types of rolling stock will be used ? (short box cars, or long flats, each determines how tight the curves can be to that industry, and the lengths of run arounds.)
4.a. In the space you have, how many run arounds will you need ?
b. will you be switching from the same direction ?
c. Will you be switching from multiple directions ?
5. How will you make up/stage your wayfreights ?
6. How many cars will your wayfreights be limited to ?( remember you will have drops, and pickups, which may vary depending on loading and unloading schedules. It is usually never one for one)
7.a. Will each industry have its own siding ?
b. Will two or more industries share a siding ?
8. How many operators will be on the layout at any given time ?
That's all I can think of right now. I'm not looking for you to post answers, only to consider them. The answers you come up with will determine your layout's design. Consider, you have one "branch" that serves four industries, one only gets a single car at at time, two get two cars each, and the last (needing coal deliveries) gets six. That branch needs to be able to handle eleven cars, and the loco, and, depending on where each industry is, what the spacing between turnouts will be. You can selective compress scenes, but you can't compress rolling stock, just use more, or less, and rolling stock determines track lengths.
I also can "see it in my head", but I go directly to benchwork, and avoid the paper. The footprint of the industry also determines where it can go, and how much track serves it. I usually kitbash industries to fit the space, or, if I already have an industry built, I use its footprint to plan where it goes ( set it on the benchwork, and figure out the best run of the siding, I use flex track, and never sectional pre curved). This, for me, works the best. After almost twenty years, I finally was able to get an "ariel view" of my modules, and was pleased to see that everything seemed very naturally positioned. Closer than you would expect, but still naturally positioned.
Pete