For a theme-first game, I use a process I call devolution, to turn that theme into actual mechanics.
Designing theme-first is like building a house. The objective is the big picture. No particular brick is going to be interesting, but the house itself will be.
In a theme-first game, there will likely not be a strong mechanical core, but more of a web of interconnected mechanics.
I don't suggest designing theme-first, for newer designers. It's more important that you get your mechanics right, than focus on the theme.
Devolution involves devolving the theme into sub-themes, and then devolving those sub-themes yet again. After a few levels, everything breaks down into actual game rules.
You want to devolve your theme into logical and obvious subthemes. When you have appropriate and realistic subthemes in your game, you can create relatable things within those themes, and they'll behave and feel like the real thing.
In Magic, there's a treasure-guarding dragon. Because the concept of a "creature" exists in Magic, the dragon is a creature. Because the concept of flight exists in the game, the dragon also flies. Because there are artifacts, they can be used as treasures. This all means that this card actually feels and behaves like a treasure-guarding dragon. It's very cool. It's not just a card with the art and name of a treasure-guarding dragon.
Let's devolve a theme. We'll keep devolving the themes and sub-themes into sub-sub-themes, until we get down to actual rules. These themes become more mechanical, as they get smaller.
Imagine we want to make a theme-first game where the players are stage magicians.
What sub-themes would you choose? Remember, you want to choose the most logical and obvious subthemes.
Here are the ones I'd choose: Magicians most obviously perform magic tricks. They also need an audience. Maybe they're trying to attain prestige. Maybe there are places the magicians can go — in a town, perhaps. Maybe the game has money and items. Maybe magicians can be injured. Maybe magicians have a level of skill. (This is probably too much stuff.)
Always remember that you're making a game, not a documentary. You're trying to evoke the theme, not replicate it. You want players to feel like magicians, as much as is possible in a board game. Players want to feel like magicians, as they imagine magicians are, even though that's probably not how real magicians are. Real magicians might spend endless hours rehearsing. That doesn't need to go into the game.
Let's now break down the "tricks" subtheme into sub-sub-themes. When I think of magic tricks, I think of things like pulling a rabbit out of a hat, card tricks, sawing a woman in half, and making an object disappear.
Let's break down the "rabbit out of a hat" trick. At this stage, we're breaking down to actual game rules — the bottom of the pyramid.
To perform the trick, you need to go to a certain place in the town, and play the Rabbit Out of a Hat card from your hand. It tells you the three steps of the trick.
The first step is to show the opponents an item card from your hand, face-down, and claim that it's a hat. Players can challenge whether it's a hat or not. (Presumably they have some clues in this regard, such as your previous behaviour in the game, whether they think you could afford a hat, or maybe whether they have one of the few hats in their own hand.) If it passes, or is unchallenged, you get two prestige points, and you can proceed to the next stage. If not, you're booed off stage, and your trick ends, for trying to pass off a bucket as a hat. You lose four prestige points.
The second step is to get the animal into the hat. This means you have to actually have an animal (a specific item card in your hand, depicting an animal.) You roll a die at this point, and add your skill score. The difficulty of this step of the trick is "9", so if your total is 9 or more, you succeed, and gain two prestige. If not, this specific failure causes you to become injured. The type of injury varies, depending on what type of animal you're trying to wrangle into your hat.
The third step is to reveal the animal, which I won't detail here.
You wouldn't want multiple tricks of this complexity in your game, and this is probably way off the actual structure you'd want your game and tricks to actually take, but this is the correct devolution process.
Let's break down the audience subtheme. They'll be coloured meeples (wooden men) drawn from a bag, and their colour will represent what trick they like. Once a player successfully completes a trick, they'll discard the meeples who liked the trick, gain one prestige and one coin for each, and draw new meeples.
I won't go into the other themes here.
Some of our sub-themes are very simple, and are directly translatable into simple rules. Money, skill, and prestige are just scores, referenced by other systems. Items are just cards you store secretly in your hand.
The rules that eventually emerge at the end of this process should be as simple as possible. They should be the minimum that's necessary to support the theme. They will likely be commonly-used rules, and they don't need to be individually interesting. You're not designing mechanics-first here.
This is why your choice of theme is so important. If you choose an abstract, convoluted, or poorly-understood theme, you can't logically break it down like this.
If you devolve your theme properly, you can create a living world, full of relatable objects. The parts support the whole. You can see this by trying to change the theme of this game. If you've done the job correctly, a re-theming won't make any sense. It's a game crafted entirely to make you feel like you're a magician.