One of the most daunting tasks for new DMs is figure out how much freedom to give the players. Campaigns operate roughly on a scale from “railroad”, where the DM confines the players to rails and sends them through a plotline with limited opportunity for choice, and “sandbox”, where the DM lets the players do whatever they want with little to no guidance. Usually the best way lies somewhere in between, where the DM decides some aspects of the campaign and determines the rest with the party during actual play.
But there’s a reason DMs have a tendency to railroad: planning is hard. If your players can visit an elven, dwarven, or orcish castle in the next session, that’s three different places to explore, three sets of NPCs to meet, and three potential sets of plotlines. If they can skip the castles entirely and go dungeon-diving or shopping, that’s even more content for you to prepare. Once the players hit a certain threshold of freedom it’s impossible to plan for everything they could do. Improvisational DMs can run a session with little or no notice, but most DMs need to know what they’re doing if for no other reason than they’re using the wonderful Live GameScreen and need all their NPCs to have portraits.
Here’s the thing, though: no matter what the players choose, you only need to plan one or two sessions. That is, no matter whether the players go to the mountains or the farms or the enemy base, they’ll meet a gruff NPC who helps them if they can prove they’re worth her trouble and they’ll be harried by the enemy archers. The specifics can change (the NPC can be a miner, farmer, or scout, and the archers can use the elevation to their advantage, set fire to crops from afar, or man enemy watchtowers) but the work you do planning the session does not triple just because the players have three options. If you create something sufficiently generic you can slot it in no matter where the players go.
That’s the basic concept behind loose session design: plan such that you don’t have to plan more. It requires a bit of making things up as you go, but you have enough of a structure that you only need to fill in the details live. For NPCs, pick a personality or motivation or occupation and leave the rest to become whatever fits the party’s situation. For monsters, build a stat block without any of the descriptive words; you know this creature has a +9 to attack and deals 2d6+5 damage, but you don’t have to know whether it attacks with a bite, a slam, or horns just yet. For maps, make something with labels but no descriptions so you can decide whether the manor’s parlor is “tasteful but old-fashioned” or “opulent and gaudy” based on whether the party visited the elderly noble or his prodigal grandson. You’re leaving intentional question marks in your session design and filling them in when you need them. Loose session design saves you time by only requiring the amount you strictly need to prepare for a session, and it doesn’t force you to improvise more than you can handle because you decide how many question marks to leave.
See, players have no idea what you’re actually doing (and, often, vice-versa). They don’t know you’re going to give them a gruff NPC anywhere they go, as long as you don’t do it everywhere they go. If your notes just say “Session 1: gruff. Session 2: helpful but ineffectual. Session 3: super racist.” they’ll meet three very different people, but you don’t have to build full stat blocks for all three of them right this second. Their actions determine whether the helpful NPC is a town guard, a lovable ruffian, or a traveling merchant. It’s the same with monsters, or maps, or even rules-intense things like magic items and rules-light things like plot threads. If you want to have an NPC task the players with stealing back an heirloom from a wizard and have him give the players a weapon in return, you can decide the wizard is a duergar and the item is an axe once the players tell you they’re heading to dwarven lands.
This works best when you design your next session with the players’ decision in mind and show them the results of their choice. You don’t want to give players the illusion of choice, you want to give them actual choice by letting their decisions have meaning. If they went to the mountains, they cut off the enemy’s supply chain for weapons and armor but the enemies at the farm might scorch the land out of spite. If they stormed the enemy castle, the enemy is in disarray but the remaining camps have time to shore up their own defenses. This isn’t about creating several branching paths or about creating a single path with a couple of mutable details, it’s about letting the players do something and deciding how the world reacts. Unlike railroading, where there’s only one path because the DM declared it, the players are guiding the story every step of the way, and unlike a sandbox campaign you’re only preparing things session-by-session instead of taking on way too much work at once.
Loose session design generally requires only two things: some item that has question marks and some resource to fill in those question marks. It’s easiest when each of these is self-contained; you decide an NPC’s personality but don’t decide their class or build, then when you decide what class they are during play you use a stat block from a book, website, or existing NPC. You can also use part of a stat block, like knowing the villain is a wizard but not filling his spellbook until you know whether he’s commanding the lizardfolk in the swamp or animating undead in the sewers. Maps with blank or incomplete keys work great, and the Internet is full of random generators that will give you names or descriptions at the drop of a hat.
This is applicable even in preparation-heavy campaigns. Consider the Zelda campaign I’m running, where most things are pretty solidly defined. NPCs have names and portraits far before they appear because I have to understand how they fit into the various quests the players can do. I have to have all my maps done ahead of time, including knowing where every treasure and puzzle is, because I have to have them ready to hand to the players when they come across the map and compass. I have to know what my monsters are doing, how they fit into the dungeon theme, how they complement or work against each other, and what their stats are so I can put them into Live GameScreen. Pretty much everything I listed above is already pretty rigid.
But I don’t have to decide what personalities or affectations my NPCs have until my players meet them. I don’t write room descriptions ahead of time so I can be as vague or specific as my players have patience for when they arrive. I don’t have to decide how many enemies are in each room, or of what types, until the players enter that room and I know how long we’ve spent on the dungeon up to then and what resources they have left. Even with all the planning I’m doing, I could be doing more, and I feel I’m better off leaving those bits until I need them. This will make things a little harder for anybody who wants to take my notes and run the campaign after me, but that’s not really my problem right now.
I think that’s the biggest argument against loose design. When you make things up on the fly it’s hard to remember what you said to the players and what you just thought about. It’s why a key NPC in this campaign changed names (twice) and why another failed to appear nearly as often as I’d intended. When my players ask a question, one of my common responses is “What did I say?”, which is equal parts “Did you listen to what I told you, or are you making up for your inattention by slowing down the game?” and “No, seriously, remind me what I said out loud because I don’t remember how I reskinned this puzzle last week.” It’s a problem that can be solved with copious note-taking, but that gets harder the longer the campaign goes and it’s never been my strong suit anyway. If you have time to make notes at the end of the session, do that. If you don’t, your players will set you straight.
Like any muscle, the more you use half-prepared, half-spontaneous session design, the better you get with it. In fact, you’re probably already doing it no matter what style you like. It’s one of the system’s assumptions, that a DM will prepare key information like NPCs and encounters but go freeform during the turn-by-turn combat and interaction. This is just taking that idea and extending it, but not so far that you reach “use the stats of a bear for every monster” levels. The point is to save DMs time and frustration while making the campaign more about and accessible to the players. The fact that it makes a DM look like a genius because like he’s sufficiently prepared for anything and everything is a side benefit.