How I Prep (Investigations and Sandboxes)
Prepping Sandboxes
I don't think there is any particular magic to how I prep my games. I have read a lot of GM advice and found a lot of it doesn't really help me all that much. I know what I enjoy doing, what irritates me, and what I can leave until the moment of play because it's already percolating inside of me. All the parts of prep that reflect doing something creative I can do easily and happily. Anything that feels like number crunching or like filling out boilerplate makes me so interminably bored that I would rather go on and do anything else due to how lethargic that kinda thinking makes me. Almost all of what I run would be considered sandbox scenarios or location based adventures so this method of prep probably only works for them. I suppose if I was making an adventure for something more linear I would just remove some of the choices and make it more focused.
Investigations
Typically I start with an idea for a problem and start to build the web of NPCs out from that central issue. I find working with people and thinking about their actions, what they might do helps me a lot more than starting with a place and putting the people in there. I don't really think about theme or anything too heady. People and their problems are so deep and engaging that it's often enough to just come up with one good one that can drive the entire adventure forward.
Once I get a solid cast of crew plus a few randoms to sprinkle in I start to approach the physical space and understand how these people and this issue inhabit the world. I have a much easier time doing this when I run Delta Green rather than Mythras or other fantasy games. Something about fantasy means I have to spend extra effort getting the basic details down to make sure there's enough for the players to feel "there." This can include looking at historical maps, pictures of actual places, or reading through novels to try and find inspiration. My legally blind self can not really draw a map to save my life so I do try to rely on digital mapmaking tools as much as I can. I typically don't show my players a map but will have them draw one or draw one for them. Mapping enough to know what the logical points of movement between locations is all the fidelity I really need.
If what I am running is a mystery it's at this point where it's mostly finished that I look at the whole and make sure the whole thing connects together in a way that makes sense and the players can follow. I don't follow the three clue rule or anything as gospel, but I carefully go through the crime/issue and make sure the NPCs and locations that it has affected are connected in ways that make logical sense and most importantly can be connected together by people who aren't me. I find it's best to thread these guiding lines at the end so the bulk of the scenario is finished and any new locations or NPCs can be added to round things out.
After that I am usually left with the busy work of statting things out and doing the data entry for Foundry. These are two parts of the process I absolutely dread but thankfully since I run D100 games I can use Retired Adventurer's genius competency bands system for most of the NPCs and save a full stat block for the most important NPCs. That method of prep is one of the ancient secrets to make BRP games really easy and fun to prep for. The drudgery of inputting stuff into Foundry though is hard to automate. Mythras has the MEG which does reduce the amount of work that someone has to do by a ton. Full NPCs can be created and randomized in moments. The parts of Foundry that chafe me are the more fiddly bits that you overlook like not hitting "Link Actors" and then all of a sudden all your NPCs are dying together or the tool's absolutely dog shit drawing tools. (I have actually bewen thinking about switching over to Sheets, TLDraw, and rolling dice at home for future games just because of how annoying Foundry can be when using a system that isn't fully supported or one that doesn't do what you want completely.) I probably use Foundry in a spartan way to most people as I only really use the drawing tools, character sheets, and the 3D dice module. All the automation and other bells and whistles aren't really my cup of tea.
Non-Investigative Sandboxes
**I need maps for all the regions, dungeons, and locations that I expect the players to go to. In a hexcrawl I will write as many hexes as I can then make very good random encounter tables that are a good mix of specific and broad so even rolls on the same result can infer something unique about the encounter ie. Luke Gearing's 2d20 bandit's post.
Dungeons though need a finished map with keys, encounter tables, and factions. Leaving the blank space in a dungeon feels cheap since it's a constructed, focused space so the same tools that hex maps use don't apply.
I don't really use a lot of tools for generation. But I enjoy reading bestiaries to find monsters or powers that could be interesting challenges or using roll4ruin to whip up a dungeon. I do swear and love the D30 tools which I find myself using a lot!.
Campaign Considerations
When preparing for a campaign I prepare things in roughly the same way but usually flip the order. I start with maps, then instead of NPCs I do factions, then figure out the most important NPCs that players will come into contact with right away. It's also the time to decide the frame of how the campaign will go which usually for my players means is it mission based sandboxes a la Delta Green and FIST or hexcrawl sandbox like Mythras and Wolves Upon the Coast. I try to get all the questions about how players wield agency and where it comes from taken care of before I get the setting into concrete details so I can design the world knowing how much the players can tear it down.
After that comes the making of name lists for NPCs, drafting up the loose details of the far away lands, and other things that are not as pressing to attend to during the first few sessions of play.