The Paid Game Master Prep System

The problem

A four-hour D&D 5e session on StartPlaying.games at five seats and $40 a seat grosses $200. After the platform’s 10% fee you net $180. That looks like $45 an hour, which is the napkin math every new paid game master does before listing their first seat. The napkin math is wrong. You are billing for the four hours the players see. You are not billing for the five-to-eight unpaid hours of prep that came before them, the forty-five minutes you spent writing the post-session recap on Sunday morning, or the thirty minutes you spent in your Discord on Tuesday answering a player’s question about their backstory. My first nine months on StartPlaying I was earning roughly $22 an hour when you actually counted the hours. The listing said $40 a seat. The reality was a 1:1.5 paid-to-unpaid ratio and a slow slide toward burning out.

That ratio is why the six-month wall exists. The pattern is brutally consistent across StartPlaying creators, Roll20 paid-game hosts, and Discord-direct GMs. Month one is exciting. Month four the GM starts feeling tired. Month five they cancel a session for the first time. Month six they drop from four sessions a week to two, and by month nine they have either quit paid hosting entirely or hit the wall and rebuilt their prep system from scratch. It is not a personal failing. It is a mechanical feature of running fresh prep for every session. The cumulative unpaid load catches up with the GM’s available hours, the session stops feeling like a paid creative outlet, and the listing goes dark. The fix is not to charge more. The fix is to compress the prep tail until the unpaid hours stay below your tolerance threshold indefinitely.

What most people get wrong

They write fresh prep for every session. The default mental model for a new paid GM is “I write this Saturday’s session this week, next Saturday’s session next week, the week after’s session the week after.” Each session is treated as an original creative project. Each one takes five to eight hours. That model survives for four months and then collapses. The fix is the opposite stance — you write your prep once, into a library, and you assemble each session from library elements with the seams sanded off. The players cannot tell the difference. Matt Mercer reuses material. Brennan Lee Mulligan reuses material. Every published adventure designer at Wizards of the Coast reuses material. The reuse is invisible because the elements are good in any configuration. Once your library has thirty NPCs, twenty-five encounters, and fifteen locations, a four-hour Saturday session preps in ninety minutes instead of six hours.

They run free Session Zeros. The community framing in most TTRPG circles is that Session Zero is a hobbyist tradition — the table sits down, talks about characters, sets the world tone, and nobody pays anyone. That works for hobby tables. It does not work for paid tables, and conflating the two silently bleeds hundreds of dollars in the first six months. A free ninety-minute Session Zero for a five-seat $35 table is $175 of unpaid time per campaign. Across six new campaigns a year that is over a thousand dollars of silent subsidy. The fix is to list Session Zero as billable session one at your normal seat rate, and to use the time to produce a one-page written agreement on attendance, refund policy, character death, tone, PvP, romance, and safety tools. The players who object to paying for Session Zero are filtering themselves out. Those are not players who will accept your refund policy, your scheduling discipline, or your safety tools either.

A working approach

The system is five operational layers stacked on each other. Each one reduces the unpaid hours per session, and the stack compounds.

LAYER 1 - Session zero as a billable service contract
LAYER 2 - The reusable prep library (NPCs, encounters, locations, hooks)
LAYER 3 - NPC stat blocks designed for paid play
LAYER 4 - Continuity system (recaps, player-visible notes, GM-only notes)
LAYER 5 - Packaging: one-shots, short series, long-form campaigns

By month twelve of running the full stack, the same listing at the same per-seat price runs four sessions a week on roughly six to eight hours of total prep, not the thirty-plus hours that same schedule cost in month one. My loaded rate moved from $22 to $48 an hour over that window. The seat rate did not change. The listing did not change. The rating sat at 4.9. What changed was the prep tail, and the prep tail is the only lever that matters.

Session zero as a billable service contract

A paid Session Zero is ninety minutes, on the clock, structured as six blocks — introductions, table rules, schedule and refund, safety tools, character and campaign Q&A, attestation. The output is a one-page agreement the players acknowledge in the campaign Discord within forty-eight hours. The bonus pack ships a ready-to-edit Session Zero contract template in Markdown with placeholder fields for your specific schedule, your Lines and Veils, and your refund schedule. The discipline is that every block ends on time. A Session Zero that overruns the ninety-minute budget is signaling that schedules are flexible at this table, which is a contract you cannot afford to start. The table rules section is the part new GMs underweight — attendance and tardiness, devices and food, tone and PvP, romance, and character death. Each one is a line item that, if not agreed up front in writing, will cost you a refund, a three-star review, or a kicked player six sessions in.

Character death deserves a sentence of its own. Four options: story death only, mechanical death with GM fudging, mechanical death with no fudging, or resurrection-available. Pick one. Write it down. Never tell a paid table you are running story-only and then secretly run mechanical-no-fudging when the dice align against a player. The discovered breach of trust is the single most cited reason for a one-star review on a GM with otherwise-strong play. Players accept whichever rule you announce. They do not forgive a switched rule.

The modular prep library in Obsidian or Notion

This is the highest-leverage chapter in the book and the layer that produces the 60-75% prep-time reduction across every paid GM I have walked through it. The library has five tiers — NPCs, encounters, locations, plot hooks, set-piece scenes — and seven tag categories: tier, theme, tone, setting, faction-class, NPC-role, encounter-type. Each tag category has a fixed, constrained vocabulary. Each element gets one tag from each relevant category. When you build Saturday’s session, you query the tags (“tier 5-8, urban, intrigue, with a noble antagonist”) and get back a list of NPCs, encounters, and locations that all fit together. In Obsidian you do this with a Dataview query. In Notion you do this with a filtered database view. In a folder of Markdown files you do this with grep. The tooling does not matter; the constrained vocabulary does.

The three reusability rules are non-negotiable: name the function not the setting (an NPC is “Grizzled Caravan Master Who Hates Authority,” not “Captain Tomas Reyes of the Sandy Coast Trading Company”), use tags not folder hierarchies, and strip the names while keeping the mechanics so the same “three goblin scouts ambush at a chokepoint with a rockfall escape” encounter runs as goblins in 5e, cultists in Cthulhu, or bandits in Pathfinder with five minutes of name swaps. The bonus pack ships the directory structure and a template for each of the five element tiers, plus three seeded NPCs, two encounters, and one location showing the tagging in practice. You import the structure once, refactor your existing prep into the templates over weeks three through six, and run sessions from the library starting week seven. By month three the library is your default prep surface and writing-from-scratch is the exception.

NPC stat blocks designed for paid play

A reusable NPC fits on a 3x5 index card. Five components, maybe eighty words: name and role, want and fear, voice and mannerism, minimal stats, and two or three hooks. The discipline is not to add a sixth component. The temptation to write a backstory and a family tree and a moral arc is real, and none of it helps you run the NPC at the table. Want and fear are the two motors of NPC behavior — want is what they spend a scene trying to get, fear is what they spend a scene trying to avoid. Voice is a speech pattern you can sustain for ninety seconds without performing. Mannerism is a physical tell — something the NPC always does — you can describe when they enter a scene. The stats block is the minimum that lets you run them: for D&D 5e, AC, HP, key ability score, save bonus, and one or two abilities. Full combat stat blocks live in the encounter library, not the NPC library.

Encounter difficulty needs a calibration warning for paid play. The standard advice in published TTRPG materials assumes a hobby table running cooperatively, which paid tables often are not. Paid tables are typically four or five players who do not always know each other well; tactical coordination is weaker than the rulebook assumes; players paying $30-$60 a seat treat a four-hour combat they lose as a refund conversation. Treat “Medium” as your default difficulty and “Hard” as the upper bound for set-piece scenes. The single fastest way to lose a paid table is to TPK them in the first three sessions of a campaign. Save the genuine threat encounters for session five and later, after the table has bought into the campaign and you have built rapport.

Continuity that does not eat your Sunday morning

The post-session work most paid GMs spend sixty to ninety minutes on per session compresses to fifteen to twenty minutes with a layered system. The input is a live notes file you keep open during the session, capturing events, decisions, and hooks as bullet points. From that single input you produce three layers: a 150-250 word player-facing recap posted to the campaign Discord within twenty-four hours, a player-visible notes page (Notion page, Obsidian Publish site, or pinned Discord post) with running NPC list, locations, quests, and faction reputation, and a GM-only notes file with antagonist plans, faction secrets, and unrevealed plot threads. The recap is your marketing surface — it is the public artifact of your campaign that prospective players read in the Discord between sessions, and the recap discipline is one of the highest-converting acquisition channels a paid GM has. A 200-word recap posted Sunday morning after a Saturday session is read by every player at the table. The same recap posted Wednesday afternoon is read by fewer than half. Decent and on time beats polished and late.

No-shows, refunds, and the table-rules schedule

A paid GM running four sessions a week sells roughly a thousand seats a year. The cancellation rate across StartPlaying, Roll20, and Discord-direct paid hosting is 4-7%. That is forty to seventy cancellations a year you have to handle without losing money, your rating, or your time. The standard refund schedule that survives contact with reality: forty-eight hours notice equals full refund or reschedule, twenty-four to forty-eight hours equals 50% refund or reschedule, less than twenty-four hours equals no refund with the seat held for replacement, GM cancellation equals full refund plus one free seat on the next session. On StartPlaying the platform enforces the schedule if you have listed it. On Discord-direct or Foundry-direct tables you are the enforcement mechanism, which means the policy has to be in writing in the Session Zero contract or you will spend Saturday evening arguing with a player instead of running a game.

The no-show protocol matters as much as the refund schedule. Session start, ten-minute hold for players who texted, then begin without the missing player. No texting them during the session chasing them down — they have the policy and they are responsible for showing up or canceling. Within twenty-four hours after, the message goes out: “Saw you weren’t able to make Saturday — everything okay? Per the cancellation policy the seat is non-refundable, and the next session is [date].” One makeup is the standard courtesy for a paid no-show. A second no-show in the same campaign removes the player. The discipline is to run the protocol the same way every time, with every player, regardless of how much you like them. Inconsistent enforcement is what makes a refund policy useless.

Safety tools, named correctly, with their designers

Safety tools at TTRPG tables are operational infrastructure for any table where a paying player might encounter content that surprises them. The canonical toolkit, in the order I introduce it at Session Zero: Lines and Veils (topics off the table or referenced without depiction, most thoroughly documented by Ron Edwards in his Sex & Sorcery writing in the early 2000s), the X-Card (any player taps to pause and reroute, developed by John Stavropoulos, and the most widely deployed safety tool in the paid-GM space today), Script Change (pause, rewind, fast-forward narrative beats, developed by Beau Jagr Sheldon, optional for default tables but useful for horror games), Open Door (anyone can leave the table at any point with no social cost, from the Larp community via Eirik Fatland’s writing on consent in immersive play), and Stars and Wishes (post-session feedback ritual, popularized by Lu Quade, runs roughly twenty minutes at session end and doubles as a feedback-gathering loop). Naming the designers correctly matters — it signals to your players that you know what you are doing and that the toolkit is not something you invented on Saturday morning.

The X-Card has one rule that GMs break with well-intentioned curiosity and the breaking of it kills the tool. When a player invokes the X-Card, the GM acknowledges, backs up the scene, redescribes without the X-ed content, and continues. The GM does not ask the player to explain. The silence is the contract that makes the tool work. The first invocation of the X-Card in a session is always awkward; the demonstration at Session Zero (the GM describes a low-stakes example, a player demonstrates the card, the scene re-describes without the spider) takes the awkwardness out of the actual use. Online tables on Roll20, Foundry VTT, Discord, or Owlbear Rodeo adapt the physical X-Card to a designated channel command or hand-raise emoji that everyone has agreed on at Session Zero.

Packaging: one-shots vs short series vs long-form

A paid game master sells three different products even if their StartPlaying listings make it look like one. The one-shot is the workhorse — a three-to-four-hour standalone session, no prior commitment required, played to a clean conclusion. The pricing power is at the lower end of the per-seat range ($15-$45), but the prep economics are dramatically favorable: a well-tuned one-shot preps in ninety minutes the first time and thirty minutes on every subsequent re-run, and the same one-shot can be sold ten or twenty times to different tables across a year. The short series (3-6 session arc, defined endpoint, $25-$50 per seat) is the bridge product — it converts one-shot players into long-form campaign players at rates around 50-70%, compared to the 8-15% who jump directly from a one-shot to a multi-month commitment. The long-form campaign (10+ sessions, $30-$60 per seat) is the recurring revenue layer. Five players paying $40 a seat weekly for six months is $5,200 of campaign revenue from one prep stream, and the library and continuity system compress per-session prep to ninety minutes or less once they are mature.

A reusable one-shot has five properties: it is system-tagged for accurate listing targeting, has a defined tone and theme in the description, reaches a satisfying conclusion in a fixed three-to-four-hour window with no cliffhanger, offers pre-generated characters or a fifteen-minute character-creation hook, and has been play-tested at least once before being listed for repeated runs. The pre-generated character path is the highest-leverage move for filling seats fast. Players who do not want to bring a character can still join. Players who want to bring a character can ask in advance. The optionality serves the broadest funnel.

Where this scales

The book covers each layer in template detail. The bonus pack ships three operational files: the Session Zero contract template (Markdown, ready to edit, every block from Chapter 2 with placeholder fields), the reusable prep library structure (directory layout and tagging vocabulary for Obsidian or Notion, plus seeded examples), and a pricing calculator (CSV with seventy-plus rows mapping session length, seat count, prep hours, and target hourly rate to a per-seat price and a profit margin). The pricing calculator answers the question every new paid GM mis-prices: at what per-seat price does my listing actually pay me my target hourly rate, given how long I currently prep? The numbers are honest. They are not the napkin math.

The mature paid-GM stack at month twelve looks like this: one library with 40+ NPCs, 30+ encounters, and 20+ locations; one Session Zero contract template reused across every new campaign; one recap-and-notes workflow producing twenty-minute post-session output; one safety tools script read at every new Session Zero; one refund and table-rules policy listed on every paid platform; three or four one-shots in active rotation; one or two short series as conversion funnels; one or two long-form campaigns as the recurring revenue layer. The work to assemble it is roughly sixty to ninety hours over the first three months of building. The work to maintain it is roughly one to two hours a week. The output is a four-session-a-week practice that earns three to five times what unstructured paid hosting earns and that does not hit the six-month wall.

Included with the book

• Session Zero Contract Template (bonus/session-zero-contract-template.md) — a ready-to-edit Markdown contract covering schedule, cadence, refund policy, table rules, safety tools, character death, and attestation. Fill in your specific fields and hand it to every new table.
• Reusable Prep Library Structure (bonus/reusable-prep-library-structure.md) — the directory layout, the seven-category tagging vocabulary, and templates for NPCs, encounters, locations, plot hooks, and set-piece scenes. Import to Obsidian or Notion in about an hour.
• Paid GM Pricing Calculator (bonus/paid-gm-pricing-calculator.csv) — a 70-row CSV mapping session length, seat count, prep hours, and target hourly rate to a per-seat price and profit margin. The honest math you needed before you listed your first seat.

Get the full picture

Readers of this also chose

Questions readers ask

I run Pathfinder 2e (or Cthulhu, or Vampire) — does this still work?

Yes. The library structure, the Session Zero contract, the safety tools, the no-show protocol, the continuity system, and the packaging strategy are system-agnostic. The chapter examples use D&D 5e because it has the deepest paid market, but every GM I have walked through this system runs at least one non-5e game and the structure transfers cleanly with minimal adaptation.

I'm a new GM with no rating yet — is this premature?

The earlier you build the system, the less it costs you. The system in this book takes 60-90 hours to assemble over three months. A new GM who builds it in months one through three will not have to rebuild it after the burnout that hits unstructured paid GMs around month six to nine. The bonus pack alone — the Session Zero contract template and the library structure — is worth implementing before your first paid session.

What if I need a refund?

Checkout runs on Lemon Squeezy. The standard refund window applies. You keep the PDF either way.

Do the safety tool scripts work for online-only tables?

Yes. Chapter 7 includes online-table adaptations of each tool for Roll20, Foundry VTT, Discord, and Owlbear Rodeo — pinned Discord channels for Lines and Veils, designated emoji or chat commands for the X-Card, round-robin or async Stars and Wishes. The mechanics are simple. The discipline is to actually run them.

How long until I see the prep-time reduction?

By week 7-12 of building the library. The first six weeks set up the structure and backfill existing prep. Sessions seven through twelve run partially from the library and partially from new prep. By month three the library is the default prep surface and the 60-75% prep-time reduction kicks in.