Player Tricks: Solving RPG Mysteries


A lot of RPGs, particularly your more skill-based, modern, and/or horror games, tend to feature frequent mysteries. The mystery might be a straight up “who/what killed this person and why?” or it might be something more abstract like “what is the villain’s plan and how do we stop it?” Basically, when the game moves beyond an up front info dump where your choices are strategic (how do we get into the encounter area and what do we use against the opponents there?), you’re often looking at a mystery.

One of the questions I’ve struggled with, and which Harbinger gives some good advice on, is what to do as a GM if your mysteries are too hard and your players can’t solve them. Sure, there are things you can do to make it easier on players, but there are also things the players can do to become better at solving mysteries, allowing the GM to step up rather than simplify her game. That is to say, I think most times that a mystery is “too hard” it’s either a failure of GMing of another type (e.g., the GM is not playing fairly with information access, is putting in too much time pressure, and/or is trying to lock the PCs into only one avenue of investigation; see this post of mine and this from The Alexandrian for ideas on how to break those habits), or it’s just that the players don’t realize the supreme power available to them in an RPG.

Not to toot my own horn or anything, but I’m notoriously good at solving mysteries in RPGs. I will routinely smash open secrets the GM thought would take all session to figure out, forcing inventive scrambling to move on to the next bit. I get pulled onto staff for live action games because I figure out the major game secrets that the plot committee thought they weren’t going to reveal for months. I really enjoy solving mysteries in RPGs.

But I’m not a mystery fiction fan, in any real sense. I don’t read many crime novels, and those I do (like Dresden Files), I don’t figure out the mystery much before the reveal. I watch a fair number of TV procedurals, but when I figure out whodunnit it’s mostly because I recognize that they cast a recognizable actor for what seems like a bit part, which probably means he or she is going to be important later. Which is to say, I’m not great at solving scripted mysteries.

What makes RPGs different? Agency.

When you’re passively consuming media, you’re limited to the information the protagonist thinks to uncover. Since the author is usually going out of the way to make sure the mystery isn’t obvious early, the protagonist will often miss opportunities to uncover information that would be very helpful to the case (or will notice it, and not remark on it until later). But when you’re playing the protagonist, you get to do things that will generate more information; you don’t have to patiently wait for the GM to dole clues out to you.

Here are some techniques for getting at that information:

Abduction, Induction, and Deduction

If you haven’t read through my article on mysteries I linked earlier, read it now, particularly the section on “-uctions.” (There’s even better explanations of them at the Forge thread I was summarizing.)

Essentially, there are three ways you can work toward solving a mystery in an RPG:

  • Deduction is the one most players are familiar with, particularly from published scenarios, where the players assemble so many clues that there is legitimately only one conclusion that can be drawn from them. The GM doles out unmissable clues as the game progresses (faster or slower depending on how aggressive you are, how your skill checks go, and whether you make bad decisions), and eventually you have enough puzzle pieces that the missing one is completely obvious. Even at its fastest, waiting for clues until you can work up a deduction tends to be really slow.
  • Induction is most useful at the mid-ranges of an investigation, because you take incomplete evidence and try to extrapolate something that explains it (but which might not be the only thing that explains it). Often, it’s the trick you use for figuring out if there are any other things you need to check before deciding you’ve got it all figured out. It’s your main way to generate falsifiable theories: we know a bunch of things, and it seems likely that X would explain them, but something else could explain them. Let’s figure out how we can prove and disprove X; if we disprove it, we need to think about these other clues in a different way.
  • Abduction in this context really means brainstorming to come up with logical explanations for clues based on known rules, which give you immediate things to check (such as whether those known rules don’t apply in this case). This is the thing you do when you don’t know much at all yet to try to figure out more things. Abduction is where RPG mysteries really diverge from scripted ones: you can jump the clue sequence pretty much whenever you want by working backwards along what seems like the simplest explanation. “The murderer got into the house somehow. One of the ways he could have gotten into the house was the nearby window. Thus, we can check to see if that’s how he got in!” Hopefully your GM has prepared enough for a lot of this tactic, because it basically means trying to skip straight to the solution using common sense and hoping proving or disproving your theories will at least narrow down the idea space you should be looking at.

I’ll drill down some more on those techniques and their corollaries.

Abduce and Accuse

While abduction is the weakest technique for proving anything at all, it’s the most powerful technique for hitting the ground running in an investigation. It means coming up with presumptive ideas, stomping around in places your character has no justification going, and being rude to NPCs by accusing them of lying and collusion. You know, the stuff you were probably going to do as PCs anyway.

This technique often benefits from avoiding a skill check. What you want here is a “yes” or “no” answer: “Is anything under the window disturbed?” “Was the butler lying about not being here last night?” “Is there anyone in town that could have befriended a unicorn or summoned a nightmare, or can we take for granted that these are actually horse hoofprints?” If the GM requires you to make a roll to see if you’d know, and you fail, then you haven’t really learned anything at all; only outright denial with a success or no roll required should be enough to dissuade you. What you’re trying to do here is not annoy the GM with fanciful ideas, but to figure out the possibility space of the investigation. Which ideas are plausible within the game world, but irrelevant to this case, and which ones actually have merit for further investigation?

Honestly, there’s a bit of metagaming involved in this technique: you can watch the GM’s face when she answers to see if it’s surprise that you’ve hit so close to the mark so fast or blankness that you’re asking about something that clearly makes no sense and was never meant to be included (GMs that really know their worlds, have prepared extensively, and/or have really good poker faces are harder to use this particular trick against).

But even beyond hopefully scoring a palpable hit on the GM, the information you get should push you closer to the right ideas. If your GM tells you “they’re definitely mundane horse hoofprints” and then later reveals that it really was a nightmare, then your GM is not playing fair; assuming your GM can be trusted to not subvert your character’s perceptions to delay a reveal, you now know not to waste more time on investigating magical horses. (This is why you want to avoid a skill check; if you fail a roll and your GM tells you something, it’s obviously a suspect answer that can’t bias your reasoning, but you’ll feel like a metagamer if you don’t let it bias you, so its best to try to avoid the risk of failure altogether. Game systems like Gumshoe move clue-finding to automatic purchases for precisely this reason; it’s generally no fun for anyone to give players the wrong information because of a failure, unless that misconception is easily corrected. Possibly with ninjas.)

The secret of abduction is that there are any number of facts about the environment that mystery fiction writers can assume their protagonists are investigating and discarding or storing away for later, but not bothering the audience with the minutia of, but as a player in a mystery game you need to make sure you have a firm “yes” or “no” on. You are probably not a skilled investigator in real life (and if you are, thanks for reading, please include your thoughts on this article and additional techniques in the comments), and even if you are, your only access to your character’s senses is what the GM describes. You can’t do the hard work of being your PC’s brain without substituting additional factual information for all the sensory information that’s not actually going from the game world to you.

Good GMs will let you make a roll to realize something your PC could have seen earlier was relevant once new information is introduced. Good players will have uncovered that information back in the original scene, written it down, and connect the dots without GM prompting.

On Avoiding Red Herrings

Proposing theories that might fit what few facts you have is a great way to generate falsifiable leads that can put you on the right track once disproven. But sometimes the GM won’t have prepared enough to make it easy to falsify them, or the existing prep will support a wild-ass theory longer than it ought to (“Yes there is an elven paladin with a unicorn for a mount in town. She’s probably not a suspect.” “Paladin, huh? What an excellent cover… for murder. Let’s go stake out her house!”).

Especially in the early stages of an investigation, you’re in danger of abducing too far. This technique should really be coming up with things for the GM to shoot down, not coming up with crazy theories and then haring off at them (unless you’re really, really good at figuring out the crazy mysteries your GM comes up with by guessing). Keep an eye out for if your GM looks uncomfortable with a theory you want to investigate; maybe she’s just mad you’re skipping ahead in the adventure further than expected, but you’re probably seeing annoyance at having to improvise something for a red herring.

There’s a school of GMing advice that supports perfect illusionism for players: if they wander off after a red herring, you try to lay things out in front of them that are interesting wherever they go, and they either eventually find their way back to the plot or they enjoy the weird ride they’re on now. (I totally discount the idea of “you move the answer so their red herring is retroactively correct” unless you’re playing InSpectres, Technoir, or otherwise have that as part of the contract; in a normal mystery game, part of the fun is trusting the GM to have an unchanging answer behind everything that’s going on that you can figure out.) In the real world of GMing, you often don’t get enough time to game in the first place, and/or not all of us are master improvisers, so the players going haring off after something false in a way that will take a lot of time and description is not always fun.

I’ll often just tell my players when they have a red herring that they won’t let go of (if subtle hints didn’t work and I’m not feeling up to playing out slack on an ultimately useless tangent). Your GM might be less lazy than me and not feel comfortable outright stopping you from wasting time for worries about metagaming or illusionism. So when you have a wild idea for a solution early in an investigation that your GM seems lukewarm or worse about, try to figure out a much simpler way of falsifying the idea before spending a lot of table time on the guaranteed test of the theory (“Before we go bother the paladin and leave the house, our ranger has really high Nature; can he tell the difference between horse and unicorn hoofprints?”)

On Shaking the Tree

If you feel stuck at any point in a mystery, and your GM doesn’t seem to be following the pulp tropes of having a man with a gun burst in on you (or ninjas attacking), it’s up to you to enact another pulp trope: go shake the tree.

This means being proactive and figuring out something you could do to try to turn up new leads, oftentimes by pissing off the mysterious villain in a way that causes her to try to kill you. The clever pulp detective knows that, if he had a day that seemed unproductive to him, but then in the evening someone tries to kill him, that probably means that the seemingly innocuous conversations of the day may have made the villain think that he was closer to her than he actually was, and they deserve further examination.

This does not mean badgering recurring and powerful setting NPCs (particularly mentors) for ideas, because usually your GM is just going to feel like you’re begging a mouthpiece for the solution. (Again, Technoir is an exception because the recurring NPCs are often also the villains and the system requires you to bother them for information.) Instead, it means revisiting NPCs and locations that have been pertinent to the case to see if a repeated examination turns up something new. In particular, if your GM has crafted a particularly hard mystery, this gives her a chance to tell you something that’s changed since the last time (e.g., an NPC that was previously being watched by an authority or didn’t take a shine to the PCs can be convinced to reveal something he didn’t before, the PCs notice something/someone at the location that could have been there by chance once but being there twice is unusual, etc.).

Even if you don’t get any new information, if you tell your GM that you’re shaking the tree, particularly if you talk to the NPCs on followup like you know more than you do, that should encourage your GM to throw some hitters at you who you can then interrogate after the fight (or search their bodies for clues, if you aren’t good at prisoners).

Induce and Improve

Abduction is good for extending the reach of individual clues, but your real meat of getting good at these mysteries is induction. This is generally the point at which you have multiple clues from multiple locations/NPCs and they don’t really speak to anything obvious yet. Induction is, in puzzle terms, like laying your pieces out on a whiteboard even though none of them connect yet, but seeing if you can arrange them in such a way that you can draw pieces that would connect them (and once you know what that piece looks like, you can go try to find it).

This is the point where it’s important to write down things you know, make sure that your PCs are not hiding vital clues from each other (deliberately or just because only one of you saw something and didn’t realize it was important), and keep an open mind. You will likely come up with a bunch of things that you think are related to the mystery, but which don’t have anything obvious to do with each other or seem to contradict. Start pitching the rest of the group ideas for things that are plausible within the setting that could explain two or more of these disparate clues (particularly things that could resolve a seeming conflict between clues).

“Alright, we know there are hoofprints outside the house that are part of our window of opportunity. We know there are really only a half dozen horses in this one-inn town, and we’ve accounted for all of them. Either someone is lying to us about where their horse was last night, or we need to try to figure out if anyone’s seen a strange horse that they didn’t think was relevant. We find the strange horse, we find the culprit, or at least someone that saw the crime.”

The goal is to come up with answers that seem to make sense and explain all the clues (or at least explain some of them and don’t contradict the others) and are falsifiable (e.g., “Maybe our villain is a crazy person and just did crazy things to throw us off” is not a valid induction unless you’re solving hard mysteries easily and your GM is now throwing you curveballs, because that’s not an induction that’s easy to prove false). You should now have some things that are either potential solutions to the whole mystery or will get you very close to the solution, and your goal is to go check them out thoroughly to try to disprove them.

You still should make a strong attempt to keep your solutions reasonable; overly fanciful solutions could still lead to red herrings, so err on the side of answers that are quickest to prove false. If your GM created a mystery where the actual explanation turns out to come down on the wrong side of Occam’s Razor (i.e., your explanation for the clues is more elegant and straightforward than the real answer), it’s now on her to figure out how to get you additional clues that lead you to the weirder solution.

Generally, if you make a reasonable attempt at inducing an answer, and your investigations prove false, the GM should try to reward you with additional information that makes subsequent inductions more robust (e.g., you shouldn’t just wind up with “Nope, nobody knows whether there was another horse in town.” but something like “I didn’t see no strangers, mister, but I do see Sir Oxney riding out to visit his mistress sometimes when his wife is out of town; he probably would’ve told you he was in bed. Don’t tell ‘im I told you.”).

Overall, while solving a mystery, you should start trying to induce early and repeat often. The group’s refrain should be, “What do we know, what does that make us suspect might also be true, and how to we confirm or deny those suspicions?”

Deduce and Destroy

If you get good enough at the other two -uctions, deduction is just a formality. It’s the way of checking your math before you go storming the prime suspect’s lair, to make absolutely sure you aren’t taking out an innocent (or someone guilty of something else). It’s a way of telling the story of what happened and seeing if it makes sense.

You need to write everything down and/or have a really good memory for clues, even more than you did during the induction phase. You’re going to lay out everything that you can prove with everything that you suspect and haven’t had disproven and try to weave a tale that explains the mystery. This is the point where you list out in exhaustive detail the five Ws and an H, and it would be the thing you accused the killer of in the sitting room in front of all the suspects if this were a murder mystery. You’ll probably just be listing it out for the other players to make sure nobody can poke any holes in the logic.

When you’re deducing, keep an eye out for discrepancies. Do you have any clues that don’t seem to fit anywhere in the narrative, or a missing W or H? It might be fine to have one thing go missing, for the final confrontation (“…and as you ran into the hills, Sir Oxley quietly assumed that it wasn’t his business what you were doing leaving the house as long as you didn’t ask him why he wasn’t at home. The only thing I can’t figure out is what kind of weapon left that hole in the victim.” “Perhaps… because you didn’t know about my trip to Numeria last year! Have you seen one of these before? They call it… a laser pistol!”). But if they seem to cause major problems with your story, you’re probably ahead of yourself and need to go shake the tree some more to get answers that lead you to a more complete narrative.

If you get it right (particularly if it was complicated) you’ll probably see a warm glow coming from the GM because you were actually paying attention and reassembled the backstory she worked so hard on.

Great Conflicting Responsibilities

Comments Off on Great Conflicting Responsibilities

This system was inspired rewatching Buffy: the Vampire Slayer. It’s intended primary for modern occult heroes or detective superheroes, but works for any game where the PCs have to balance a normal life (including school or a day job) with the need to investigate in order to find and stop opponents. Virtually all scenarios should involve enemies that grow in power and get further towards fruition of schemes as time passes, granting big rewards to the players for constantly working to curtail their activities, get wind of their plots, and quash their plans early.

The examples below are system-agnostic but assume something with difficulties on roughly a ten-point scale and low-granularity experience points (like oWoD). Adjust values accordingly for other systems.


All investigation attempts take an hour or two and can include:

  • Patrolling: Both superheroes and monster-hunters tend to get their first leads by running or flying around the city looking for heads to crack and vampires to stake. In addition to keeping an eye of the streets for anything big or weird, this tends to reduce the number of minions available for bigger capers.
  • Research: Less formidable characters can keep an eye out for upcoming occult junctures or attractive targets of crime in order to get a clue that something might go down soon. Once someone has a name or description of a threat, research involves cracking books, trawling the internet, or hitting up periodicals looking for patterns, secrets, or weaknesses.
  • Forensics: Sometimes, the villains leave a crime scene that our heroes can get to (ahead of or with the blessing of the police). Going over the scene can yield clues, as can taking away any material or mystic traces left behind for evaluation in the lab.
  • Gathering Information: Sometimes, your more gregarious characters can get word that something is up by keeping up with contacts. Once a threat has presented itself, hitting up known informants can be the best way to find exactly what’s going on and where it’s going on at.

Depending on how you like to run mysteries, you can either give out fixed successes based on relevant skill totals every time a player takes an investigation phase or have players make rolls and track margin of success for relevancy. You can track accumulated successes toward a conclusion where they know everything they need to pursue the endgame or have various pieces of information available to various types of investigation with the players trusted to decide when to act upon them. The important thing is that investigation is a time-consuming process that feels like building up information toward a goal rather than just following pre-scripted encounters.

In the background, the villains should always have their own progress bar toward some goal. Patrol might set their progress back by defeating minions and capturing materials, but ultimately their plan is proceeding toward some hidden end in an unknown place, and the job of the players is to ascertain both in time to stop it.

Each day, every player character gains one free “investigation point” that can be spent to:

  • Make one attempt at patrolling, research, forensics, or gathering information
  • Train non-job/school skills (see below)
  • Lower either Stress or Delinquency/Dereliction by one point (see below)

This represents using free time to pursue the investigation, train, or catch up on relaxation or work.

Additional Points:

  • Each player can choose to gain one additional point per day by taking on either a point of Stress or Delinquency/Dereliction. This represents either staying up late for another round or cutting class/skipping work for a couple of hours.
  • Each player can choose to take up to two more points, but each point past the second represents majorly ditching out of school/work and the stress this entails, essentially spending all day on extracurricular activities.
  • On weekends, the GM may choose to just award three points for free (with the fourth point available for a single point of Stress or D/D, representing the stress of blowing off a whole day of free time or not doing homework).

Needless to say, most villainous plots should proceed fast enough that the PCs won’t be able to stop it with just the one free investigation point each day. The point of the system is that stopping the bad guys involves having to make cuts to free time or slack off at school/work.


Players have to spend investigation points (on a one-for-one basis) to spend experience points on any skills that can’t be justified being learned from normal school classes or on-the-job skills. If you want to get that 4 exp upgrade to Getting Medieval, you need to spend time on weapons training that you’re not spending on investigating. Training is a major downtime activity, ensuring that players may not totally zero out Stress and D/D between stories (but also see Long Downtimes, below).


Stress represents exhaustion, lack of concentration, and just general frustration at spending all one’s free time on the mission. Stress becomes the minimum difficulty for all rolls. In a system like Unisystem with a fixed DC, your stress total is similar to an opposing roll on every task (i.e., stress grants a success penalty equal to the margin of success it would achieve if it were a roll on that result). The intention with either version is that Stress shouldn’t become much of a problem until it gets fairly high. Players should be tempted to throw some points into it for extra investigation points because it’s not a big deal… until it is.

Stress has a practical cap at the maximum reasonable difficulty for the system (or the result of a really good roll, for fixed DCs). At this point, the character is so exhausted that even the simplest tasks are huge efforts.


Delinquency represents skipping classes at school, while Dereliction represents taking long breaks, getting in late, or leaving early at work. Both are the kind of thing that eventually get you in a lot of trouble. A student whose Delinquency reaches the same number as the practical cap for Stress is visited with whatever punishments seem warranted (suspension, detention, or even expulsion, plus likely grounding by parents). An adult whose Dereliction reaches this number is fired. Additionally, it works like Stress to set a minimum difficulty for all interactions with school officials and parents (for students) or employers (for adults); since Stress is a minimum difficulty for ALL rolls, it takes precedence if higher. Once it gets fairly high, the GM may initiate scenes with the PC having to talk officials, parents, or employers out of assigning more onerous tasks, with failure resulting in some responsibility that will gain an additional point of Delinquency/Dereliction if skipped.

Students can take a trait called “Honors” that represents being good at school and having easy access to school resources like the goodies in the science labs. Adults can take a trait called “Income” which works like wealth traits do in any system. Both of these traits are “free,” but essentially set a higher starting value for Delinquency or Dereliction (e.g., if you have Income 4, two points of Dereliction raises you to 6). The students with the brightest futures have more onus on them to live up to expectations, and the adults with the best jobs have more people that will notice if they skip out of work too much. These traits should scale so their maximum is about half the cap for Delinquency/Dereliction.

Players can purchase levels of “Gifted” or “Idle Rich” with character points as normal advantages, representing access to Honors or Income without the associated responsibilities. For example, if you have Gifted 3, you could choose to have a total Honors of 5 while only starting at 2 Delinquency.

Long Downtimes

This system assumes that there will be fairly limited downtimes. Stories represent an active few days or weeks, and then the next story starts only a week or two after the last one. In this case, there are no need for modifications; players will use the time to buy down Stress and D/D earned during the last story or spend points on Training, but will probably not have time to accomplish all their goals before the next story starts unless they ended the last one with very low totals.

If your game includes longer downtimes, simply allocate as many points as they spent on training minus 1d6 to their choice of Stress or D/D. This represents other life stuff coming up; either adventures too minor to note, or home events that made a nuisance of themselves.

Writing and Solving an RPG Mystery

1 Comment

Originally posted February 2009

Last year I found a good thread on writing mysteries for RPGs. The advice was for games that follow a model more typical of crime shows (investigators find crime, make theories, test them, make new theories if wrong) than of the typical RPG model (investigators find clues, follow them to next scene, find more clues, etc.). Below are my summarized notes on the thread, rephrased and expanded a bit with my own thoughts.


  • Deduction: Given a known case and a rule related to the case, the result of the case can be determined. (Given that unknown beans are from a bag, and that all beans in that bag are white, the unknown beans must be white.) This is the strongest type of investigation, but requires most of the information before reaching a solution.
  • Induction: Given a known case and the result of the case, the case’s rule can be determined. (Given that these beans were taken from the bag and all of these beans are white, all beans in the bag must be white.) This is the second strongest type of investigation, because reaching a solution generally allows the solution to be further tested.
  • Abduction: Given a result of a case and a rule applicable to the case, the case itself can be determined. (Given that these beans are white and all beans in that bag are white, these beans must be from that bag.) This is the weakest type of investigation, because it proves nothing, merely suggests a solution that can be tested in other ways. However, since investigations will often start with a result (the evidence of the crime) and rules (forensic sciences) but will not initially include the case (the method and motive of the crime), abduction is the standard investigation method.

In order to strengthen abduction you can turn to deduction, induction, or comparison:

  • Deduction: If the abduced case is true, using other rules what other result must be true? (If the beans are from this bag, given that there are a standard number of beans in a bag, the bag must be missing this number of beans.)
  • Induction: If the abduced case is true, using other results what other rules must be true? (If the beans are from this bag, and the bag currently contains a certain number of beans, these beans plus the beans in the bag must equal the original number of beans in the bag.)
  • Comparison: If the abduced case is true, no other adduced case can adequately explain all the results. (Given that these beans are white and all beans in that bag are white, are there any other bags of white beans from which these beans could have originated?)

Given all the facts (clues) of the result, what case might have been true to exactly produce all those facts?

Mystery case blocks/stages:

  • Stage 1, The Scene of the Crime: Characters investigate the result, finding all details however trivial. (Designer must place all essential clues to describe the result and ensure they are discovered.)
  • Stage 2, Abduction: Characters discuss, adducing a small number of cases that would explain the result and depend on certain or likely rules. (Designer encourages them to find as many solutions as possible and grade them on how well they make the known fact predictable, and encourages them to find the simplest hypothesis.)
  • Stage 3, Investigation: Characters investigate possible violations of the adduced cases, eliminating rules that turn out to be unsupported by additional evidence, and looking for additional, non-obvious clues that would support a particular case. (This is the major part of the game, and lots of new evidence will be uncovered, some of it superfluous. At the end of the block, the characters should be very close to a correct case.)
  • Stage 4, the Case: Characters rebuild the remaining cases until they can eliminate all but one, repeating the third block until only a single, watertight case remains. (Designer should make it hard to get to block 5 with an incorrect case, instead funneling the characters back into block 3 if there are still holes. In murder cases, killing a suspect or another murder in a way that gives a suspect an alibi is the usual method for breaking an incorrect case.)
  • Stage 5, the Plan: Characters develop a plan of action that will confront the culprit of the final case in a way that proves the case and produces a climactic scene.
  • Stage 6, the Climax: Characters deal with the fallout of the fifth block, possibly using previously determined information to fight or track a culprit that has fled after the revelation of block 5.

Episodic blocks:

  • Teaser: Introductory material and first beat of the discovery
  • Act 1: Block 1, likely ending on a hard-won clue
  • Act 2: Blocks 2 through 4, likely ending on the revelation of the case
  • Act 3: Block 5, likely ending on the villain raising the stakes
  • Act 4: Block 6, final confrontation with the villain/climax and denouement

Four or five major beats per act.

Inventing the Case:

Step 1: Straight Story

Who – Who did it and who was hurt?
What – What was the crime?
When – When was it carried out?
Where – Where was it carried out?
Why – Why was it carried out?
How – How was it carried out?

What, Where, and the second Who should be immediately obvious, with the When following shortly unless concealing the time is important (and a range of times should still be apparent). The How should become evident during the intial investigation. The first Who and the Why are identified by the investigators.

Step 2: Messiness

What clues were left behind as a result of the crime not being perfect? Advanced: were there any coincidences that covered up clues?

Step 3: Bystanders

Who was connected to the victim and can provide information relevant to the crime? Who gains and who loses and who knows about it? Does he have enemies or people that feel strongly about the crime?

Step 4: The Result

What is the exact timetable of events? Who was there and who was nearby? What basic clues were left, and how are they to be found? These are essentially points that the police will likely find with a few hours, if they take an interest.

Every NPC that is connected to the victim or location but not involved in the crime should have a routine or an alibi that eventually eliminates him or her as a suspect. Any actions the criminal took to distract the NPC should become obvious as a deviation from the NPC’s intended routine.

Step 5: The Difficulty

Why can’t the case simply be left up to the police to solve?

Step 6: The PCs

How do the PCs get involved? If they’re called in, who calls them and why? If the person that called them in has a theory as to what happened, it’s probably wrong.

Rule of 7:

No more than seven new named NPCs, important clues, or major ideas. Seven key concepts total is even better.

Edit (7/2010): Rob Donoghue has some excellent advice on different types of murder mysteries that syncs nicely with the information above.