I think the last thing I should have done differently with my Fallout campaign would have been to allow the players to guide the campaign. The Fallout video games provide you with problems (like "find this technology" or "make your way to this settlement" or "earn some money to get better equipment") but then allow you to go anywhere you want. You can decide what problems you want to solve, and in what order. You can choose different approaches, ally yourself with different factions or individuals.
I had ambitions to do that, but never followed through. For the most part, I set up a linear series of missions for the players to work through. I did craft those missions based on the things the players seemed interested in, and I always tried to be accommodating whenever the players wanted to take an unexpected sharp left turn, but still, it wasn't anywhere near the open-world sandbox experience I had originally considered.
What I should have done was populated my map with allies, enemies, and resources, and then confronted the players with a collection of problems and opportunties, and let the players decide where to go. It would have meant more up-front setup time, but it would have saved me some between-session prep, as I wouldn't be planning out missions one at a time, I'd just be presenting the world as the players encountered it.
A sandbox approach won't be useful for all my future games; many games, like Monster of the Week, are supposed to be a series of self-contained episodes. But I would have been more faithful to the Fallout concept with a sandbox game, and though I have a nicely-detailed world now that the campaign is drawing to a close, I would have rather had this big pile of people and places ready from the start.