Many people think D&D just cannot be played in under 8 or even 4 hours. Yet I've been running one-shot campaigns with friends and at local conventions for the past two years, in about 4 to 2 hours per game. Look, D&D can be laborious if it's always only an epic. Sometimes D&D needs to be episodic, bite-sized, manageable. Running one-shot campaigns makes D&D, in a sense, a board game that you can take out of the closet and play when everyone else is ready to default to a quick game of Cluedo or Monopoly or any of those games that no one really likes but tends to play anyway, out of habit.
So I encourage people to play D&D without the burden of grandiose story arcs or persistent characters and loot, and as I've been running one-shots so often lately, I've collected some notes on what tends to work and what may not. Obviously all of these notes are subject to your own DM style and your players, so this isn't meant as a manifesto. These are observations (and probably the designers of one-shot systems like D&D Adventure League and Pathfinder Society have know them for years). Read them over, evaluate them, modify them, and test them out.
If you start your players receiving a quest from someone in a town, then your players are going to assume, quite fairly, that it's part of the game. That means that your players are going to question the quest-giver. They're going to investigate the town, they'll try to hire people to help them on their quest, and purchase supplies, and plan for their journey. And then 2 or 3 hours into a 3 hour game, they'll be ready to actually go on the adventure.
Start your one-shot at the entrance to the dungeon (assuming you've prepared a dungeon), or the dungeon-equivalent. Let the prologue be a prologue: tell the players their mission (if there is one), and then start the game part of the game so that if the players want to spend an hour turning over every stone, you can at least justifiably spring some traps on them rather than coming up with mundane items for them to discover while second guessing a quest giver who they'll probably never speak to again anyway.
If you can't do a cold start, for some reason, then one trick I've learnt from some Pathfinder Society campaigns is to make sure that the quest giver is a person of authority of the player's faction or organization. In official Pathfinder Society adventures, there's an in-game organization called the Pathfinder Society, and players are each members of the society, so when they begin an adventure, it's because a Venture Captain within the organization has requested them to go on a mission. The PCs trust the Venture Captain because they understand that this NPC is a figure of authority within their ranks. If a first-time Pathfinder Society player doesn't know this and starts to interrogate or suspect the quest give, it's easily explained: "Your character knows and respects the Venture Captains of the Society".
Your setting may not have a Pathfinder Society, but it's easy to invent a faction or guild, or any generic "trusted friend" who can bestow a quest upon your players. Reminders that they know and trust the NPC can be generously applied as needed.
To prevent the players from spending literal hours in town preparing for the quest, the quest giver should also have arranged transportation to the adventure site, and that transport is leaving NOW. Whether it's a boat or a caravan or a [cowardly] guide (if the guide isn't cowardly, the players will attempt to recruit them), it has to be leaving within minutes of the quest being given to ensure that the next stop of the game is...the game.
D&D has infinite possibilities, and that's why we love to play it for days and days on end. Dungeons, however, are limited in scope, and that's exactly what makes them good for a 2 to 3 hour adventure. In a dungeon, there are a finite number of things for your players to explore, and if they're lingering too long in one place, you can always have a wandering monster drive them deeper into the dungeon.
For one-shots, an actual dungeon is better than a dungeon-like forest, or a dungeon-like town, or any other "it's basically a dungeon", because once the players are in an actual dungeon, there's usually no negotiation. Nobody ever proposes leaving the dungeon to rest or get more supplies or hunt for food, but if your players are in any other environment standing in for a dungeon mechanic, then they often feel free to wander back out of it.
It sometimes helps to remember that early D&D very often didn't have any story other than the story of the dungeon itself. Although modern sensibilities often expect (or you think they expect) a quest and a reason for characters to exist, players do know they're playing D&D. They understand that the reason they are playing the game is because they want to play the game, so they don't actually need justification.
If you're trying to teach a new player how to play D&D, there's a strong argument against starting with a character build. I'm a huge proponent of teaching new players how to build characters, and frankly I consider building a character both a fun mini-game and a valuable look into how the game works. But I don't think it's material appropriate for the first lesson. It didn't used to be that way, but building a character in the 21st century isn't what it once was. You can spend an hour building a character in Pathfinder or 5e, so do that for your players in advance.
The danger is that players don't feel any connection to characters they didn't build. To combat this, I leave a few things blank on the character sheets and let the players fill them in. These are:
Name: Some players love thinking of their own character name, others can't think of one, so I let them decide whether they want to invent something, roll on a d100 name table, or use an online name generator.
Alignment: I explain alignment, and let the players decide on one for their character. This is also a great way to explain to new players that they'll be working together
Languages: Often, a character ends up with the ability to speak in one or two extra languages. I never choose these as I build the characters for a one-shot, and instead let the players decide during the game.
Mechanically, the effect is, essentially, a natural 20 on translation of pretty much whatever language you throw at your players, but rare is the occasion that I put a language into a one-shot that I don't want my players to be able to decipher. Letting the players choose a language is an easy way to connect with their character. Giving my party "free" translations is a small price to pay for player character investment.
If you're not sold on excluding the character build, or else you have players who really want to experience a character build, set the PHB or Core Rulebook aside. Normally, I'm an emphatic proponent of going to the source, reading the manual, and learning by doing. But if you've only got 3 or 4 hours, the rulebook is only going to slow you down with details you just don't need.
Here's my quick build workflow:
Pick a class and race. Look up relevant attribute and proficiency bonuses, then literally set the rulebook aside.
Use the standard array (15,14,13,12,10,8) and distribute among attributes.
Grant a reasonable proficiency bonus: +2 for low levels, +3 for mid, +5 for high.
Select 2 saving throw proficiencies.
Select 3+level skill proficiencies, based on what makes sense for a character or else based on what the player wants.
Select 1+level cantrips and 1+level spells and equipment.
I like playing at level 1. There's something refreshing and comforting about that fresh player character smell, and the sense of discovery and developing a character. But a one-shot implies that characters aren't going to level-up, so playing a level 1 character isn't all that meaningful.
Generally, I don't ever see a good reason to start at level 1. A level 2 character is a great starting point for new players: it's more exciting than level 1 but not as complex as level 3 and 4. Even for experienced players, I favour low-level characters (2 to 5 or so) only because high-level characters expect that the player has leveled with the character and therefore has had time to learn the abilities and nuances.
For new players, the concept of a cooperative game can be surprising, and for experienced players, the idea of dropping into a PC that they haven't built themselves can be jarring. It can help both types of players, and serve as a prologue for the adventure itself, to spend a few moments establishing how and why player characters know one another.
It doesn't have to be an elaborate back story, it doesn't have to have any emotional impact, but establishing why characters know one another helps the players decide early that their party trusts one another, would never steal from or threaten one another, and so on, and also provides guidance to how to role play the PCs.
A caveat for both a one-shot and an ongoing game: if a particularly creative player conjures up an elaborate back story, a good DM acknowledges and validates it during the game. You're probably not going to develop a player character's story arc in a one-shot, but just acknowledging the effort put into the back story ("You get the sense, because you yourself were an outcast for so long during your youth, that this creature is not evil, but is actually trying to communicate...") makes all the difference from a player's perspective.
One-shots aren't like big campaigns. You don't have to plan much, and in fact for a short session, the less you plan, the better. Draw three rooms on a napkin, grab your Monster Manual and some treasure tables, and start playing. Your players will fill in for the gaps in your content: they'll set off traps, they'll find secret passages, they'll get themselves injured and infected and poisoned, and by partaking in the dungeon they will generate the story.
If you've taken my advice and set your adventure in a dungeon, you can do even less than that, thanks to the Mixed Signals Dynamic Campaign Resource system. You can show up to a one-shot with some prebuilt characters and some Mixed Signals tables, and generate the dungeon literally as your players move through it. Zero prep work, infinite (but contained!) fun.
Bring players and their characters together by emphasizing when one player character has a significant impact on the life of another player character.
Don't make a big deal about it. Just make mention in passing when one player has helped another.
"Yes, you detect a trap. You guys, she might have just saved your lives, because Kargrall's foot is hovering over the trigger-point now."
"Good thing Nekki used a healing spell in the last room, because the orc's sword slashes across your chest for 8 points of damage."
And so on.
One-shots are most fun when the players get lots of loot, lots of magic items, and lots of surprises in general. The problem is that most players are afraid to use what they have for fear that they'll need it later. It doesn't matter the situation, there's always a time they'll need it more later.
To avoid this, I emphasize that they must use what they find, or else lose it before the end of the game. How the players lose an item depends on the item. Maybe the magic just fades, or maybe a cool weapon starts to suddenly rust, or maybe it gets stolen mysteriously. The point is (and I sometimes make this explicit by saying "There's a note on the item, reading this item shall return to dust if taken from this dungeon"), items found in the dungeon are meant to be used within the dungeon.
Warning If you don't make it explicit that items become useless if taken from the dungeon, then you should hold on to the character sheets at the end of the game, or else deface the character sheet ("This character has RETIRED") so that your players don't come back to play more, armed with their severely overpowered character.
D&D is best when it's fun. Of course, "fun" is defined differently by every individual, but as a DM you likely understand that. The point of a one-shot is to experience the most fun parts of D&D, so don't shy away from that. Use it or lose it!
Map image by Mixed Signals. Creative Commons BY-SA.