It's 9:47 PM. Your party has been interrogating the same tavern keeper for forty minutes. They missed the bloody footprint in room three, didn't roll high enough to notice the signet ring, and now they're arguing about whether the bard "sensed a vibe" from a guard they've never met.
You know who did it. They don't. The session ends in thirteen minutes.
Sound familiar? If that 11 PM panic is becoming a regular feature of your sessions, you're not alone. Running mysteries is one of the hardest things a GM does. You're balancing hidden information, player agency, and pacing all at once, and one missed clue can tank the whole night. So GMs default to railroading: funneling the party toward the "correct" suspect, spoon-feeding clues when things stall, or worse, just telling the rogue what they "would have noticed."
There's a better way. And it doesn't require you to be Agatha Christie.
Why Mystery Adventures Fall Apart (and What Actually Causes It)
Here's the core problem, as GM designer Justin Alexander put it in his original essay on the Three Clue Rule: mystery scenarios in RPGs fail because every clue is actually four failure points stacked on top of each other.
Your players have to:
- Choose to search the right place
- Recognize the clue as important
- Succeed at whatever roll you're gating it behind
- Actually connect the dots to the right conclusion
That's four coin flips per clue. Even with a generous 75% success rate at each step, the odds of getting all four right land around 32%. No wonder your rogue keeps failing Investigation checks at the worst possible moment.
Most published mystery adventures are written like novels, with a single chain of revelations. Miss one link, and the whole chain breaks. That's where the panic-flips-to-railroading instinct comes from. You didn't want to hand the answer over. The adventure structure just didn't leave you another move.
What Is the Three Clue Rule? (And Why It Keeps Investigations Alive)
The fix is embarrassingly simple. Justin Alexander calls it the Three Clue Rule: for any conclusion you want the players to reach, plant at least three clues pointing to it. (He's since published an expanded version of the rule if you want to go deeper.)
Not one. Not two. Three.
Why three? Because your players will miss one, ignore a second, and misinterpret the third, and that's when the investigation actually gets fun. They'll swear they solved it themselves. They'll argue about which clue was the "real" one. They won't notice that you padded their odds.
Here's a question that makes clue design click: for every suspect, do the players know whether they had a motive (a reason to do it), means (the ability to do it), and opportunity (access to the victim or scene)? Every clue you plant should be answering one of those three questions, or throwing a false answer at one. The Angry GM points out that most mystery failures come down to investigators having a suspect they can't quite rule out, not because the mystery is complex, but because the GM never decided whether the opportunity existed. So when you're stuck on what clues to prep, ask: which suspect is missing motive, means, or opportunity coverage? Plant clues there.
A Worked Example: Nine Clues Across Three Conclusions
Let me give you a concrete example. Say the killer is a werewolf butcher named Danner, and you want the party to connect three dots: it's a werewolf, it's someone the victim knew, and the butcher shop is key.
Instead of planting one clue per dot, plant three:
It's a werewolf:
- Oversized claw marks on the doorframe
- Wolf tracks that start as paw prints and become boot prints
- A silver locket the victim was clutching
It's someone the victim knew:
- Love letters signed only with a "D"
- A diary entry about "dinner with him tomorrow"
- A witness who saw them arguing at the market
The butcher shop:
- A crate stamped "DANNER'S MEATS" in the alley
- Blood traces that don't match the victim's
- A day-planner entry: "deliver order, 9 PM"
That's nine clues for three conclusions. The party will find four or five. They'll feel like detectives. You'll feel like a genius. Everyone wins.
How to Prep a Mystery Without Writing a Script
Here's the mental shift that changes everything. Mike Shea of Sly Flourish puts it plainly: build situations, not mystery novels.
A plot is a chain of scenes that happen in a specific order. A situation is a set of facts, people, and places that exist whether the players investigate them or not. The villain has a plan. The victims exist. The clues are scattered. What happens next depends on what the party does.
This means your prep looks different. Instead of writing "Scene 3: The Party Visits the Docks," you write:
- Timeline: What's already happened, what's happening now, what happens next if nobody intervenes
- NPCs: Who knows what, who's willing to talk, who's hiding something
- Locations: What clues exist in each place, not which clues the party must find
- 10+ portable clues: Information you can drop wherever the party happens to look
That last one is the trick Shea hammers on. Keep a running list of clues that aren't glued to specific locations. When your players decide to question the baker instead of the guard captain, you can pull a clue off the shelf and hand it to them through the baker. The players don't know you did that. They just feel like their choices mattered.
Honestly, this is where I used to get tripped up. I'd spend three hours writing a perfect scene at the docks and then the party never went to the docks. Now I spend that time writing clues, not scenes. The scenes build themselves at the table. (If your campaign notes are still scattered across six Google Docs and a sticky note, fix that before your next mystery arc. A portable clue list only works if you can actually find the list.)
The Most Radical Version: Don't Decide Who Did It
Here's a design move worth knowing about even if you never use it directly. Brindlewood Bay, Jason Cordova's cozy-murder game from The Gauntlet, simply doesn't decide who the killer is before the session starts. The GM preps a list of unanchored clues — "a heated argument about an inheritance," "an unusual stain on the victim's coat," "a witness who won't meet your eyes" — none pinned to specific NPCs. The GM drops whichever ones feel right as the session unfolds. At the end, players formalize a theory using a "Theorize" move, and that theory becomes what actually happened. The killer is decided by the players, not the GM.
You can't railroad anyone toward the wrong answer when you don't have a right answer prepared. D&D GMs can steal the principle without buying the game: prep ten portable clues, keep the killer ambiguous in your notes, and let the party's strongest theory become canon. It's terrifying the first time. It's freeing every time after that.
When Should Players Roll to Find a Clue?
This is the question that splits GMs into camps, and I've got an opinion.
The GUMSHOE system from Pelgrane Press, designed specifically for investigation games like Trail of Cthulhu and Night's Black Agents, takes a hard stance: if a character has the relevant investigative skill and they're looking in the right place, they find the clue. No roll. The dice are for interpretation, for action, for the interesting choices, not for gating core story progress behind a d20.
Robin Laws, who designed the system, calls this an anti-rule — a rule that exists not to determine what happens, but to give the GM permission to not roll. His observation is sharp: even GMs who intellectually know they shouldn't gate critical clues will do it anyway, because "like a gun on the table in the first act of a play, if you introduce a rule, it's going to go off." You can write your own anti-rule on your DM screen tonight: if the party is in the right place and looking, they find at least one clue. That sentence is your permission slip.
D&D doesn't work that way by default. But you can steal the principle. Here's my rule of thumb at the table:
- Finding a clue that unlocks the main plot? No roll. Just tell them.
- Finding bonus information that makes things easier? Roll for it.
- Interpreting what a clue means? Roll for it, or even better, let the players argue it out.
- Acting on the clue (breaking into the shop, tailing the suspect)? Definitely roll.
I tried this about two years ago in a Curse of Strahd run and my investigation sessions got maybe 40% faster. The party stopped feeling punished for bad dice. They started feeling like detectives who were bad at parkour.
What to Do When the Investigation Stalls
Even with the Three Clue Rule, sometimes the investigation just... stops. The party is staring at you. Nobody knows what to do next. The silence is getting weird.
Do not, under any circumstance, read your notes aloud.
Instead, make the villain do something. This is the single most important technique for running mysteries. When players are stuck, advance the clock. The antagonist isn't sitting still waiting to be caught. They're:
- Killing the next victim (now there's a fresh crime scene)
- Covering their tracks (now there's an NPC acting suspicious)
- Sending a warning to the party (a threat, a bribe, a fake friend)
- Executing the next step of their plan (a robbery, a ritual, a disappearance)
Each of these gives the party a new entry point into the mystery without you having to break the wall and feed them information. A second murder is a gift. A bribed city guard tailing them is a gift. You're running a world where the bad guy has agency, and the players feel the pressure instead of your notes.
I had a game master in our Discord mention she ran a mystery where the party got so stuck she had the villain mail them a mocking letter with a literal wax seal of his crest. They were furious. They traced the seal. They had the killer in two sessions. She calls it "the arrogant villain maneuver" and it's become my go-to when I sense a stall coming.
A related note: if your investigation is stalling because the party is more interested in starting bar fights than chasing leads, the villain-clock technique gives you a way to redirect that energy toward the plot. Our framework for handling chaotic players goes deeper on that without the lecture.
How Complex Should a Mystery Be? Keeping It Small Enough to Actually Solve
One last thing. If your mystery requires the party to connect seven suspects, three timelines, and a political conspiracy with four noble houses, you are writing a novel. Novels are read alone, slowly, with the ability to flip back.
Your players have been drinking and making dick jokes for two hours. They will not remember the name of the third courier.
Keep mysteries to three suspects max for a single-session whodunit. Keep the central truth to one sentence you can say in ten words or less. If you can't summarize the solution on an index card, your players won't reconstruct it. Shea's advice here is dead right: simple enough that players can actually piece it together.
Save the seven-suspect political conspiracy for a multi-session arc, and even then, put a session-zero-style recap at the top of every game. Your session zero checklist is a good starting point, but for ongoing mysteries I keep a single shared doc titled "What The Party Knows" and update it after every session.
Your Three-Clue Challenge: A Homework Assignment for Your Next Session
Here's your challenge for your next session. Pick the next big reveal your players are working toward. Write down three clues pointing to it. Three different clues, in three different locations, using three different senses or approaches (a visual trace, a witness statement, a physical object).
Drop one this session. Hold the others in reserve. If the party finds the first clue, great. If they don't, next session an NPC happens to mention something. The session after that, a body turns up with the third clue on it.
You just built a mystery that cannot break. And you didn't railroad anyone.
What's the worst a mystery has ever gone at your table? I want the real horror stories. The four-hour interrogations, the red herring that became the main plot, the time the bard seduced the actual killer in session one. Drop them in the comments or in our Discord. I'm collecting them.
Run Your Next Mystery Inside ScriptoriumGM
If your portable clue list currently lives in a Google Doc, a Notes app, and a napkin, try running your next investigation inside ScriptoriumGM. Drop your nine clues into the Campaign Library, tag them by suspect and location, and when your party swerves toward the baker at 10 PM, your AI Campaign Assistant can surface the clue that fits — in the baker's voice — without you rifling through tabs. The Three Clue Rule does the design work. We do the "where was that note again?" work.
Start free and bring your next whodunit.