Typically from a visual idea that emerges in my head, or the fragment of a story, some vague idea. I often do architectural concept sketches. For example, Penny Dreadful 2 came from this sketch: My first Thief mission was made to plan, and only changed slightly. The second and third were almost completely improvised. The last was built around sketched architectural elements, but the goals and structure of the mission changed significantly. And I have a bunch of sketches and maps for missions that haven't been realised (yet). There are similar patterns in TDM: PD1 and Return to the City were fully improvised, PD2 and PD3 built around concepts, while Fauchard Street and The Rats Triumphant were speed builds with precise objectives and relatively little organic development due to time constraints. I have tried multiple building styles, including prefab-heavy missions and others. They have their respective draws and problems. For PD3, I am shying away from too many prefab elements, since I want it to have a very handcrafted, non-repetitive look that's mostly BSP-based. In most missions, I build scene by scene (although I often return to older scenes), and don't really enjoy the stepwise approach of doing a level task by task. I like to exploit opportunities that crop up during the building process, and use them to build new areas. If there is a place that looks like a fun location to climb up, why not make it possible? I procrastinate a whole lot, or set a project aside for a long time. Not a great solution, but that's it. There is also a time when I start to lose interest in a project, which is a sign telling me to wrap it up and release it. This is also why I'm reluctant to do mission updates. Gameplay trumps realism, story, and really every other concern. Mood is also very important. Light and sound contribute much to mood. I'm all for testing the mission as much as possible, with a lot of different approaches - stealthy, ghosting, violent, no equipment use or lot of equipment use, and so on. It makes the mission smoother, and it is a good source of new ideas. I am against obscure secret switches, since they can really break the flow of a mission. Generally, core objectives should not be too hard or too obscure - but they are fun to leave as extras. Learn to avoid invisible walls, excessive linearity, and other restrictive trappings of modern gaming. Enable, and don't constrain the player. Think in terms of alternate solutions to gameplay challenges. Allow yourself to be surprised. Performance-related problems, due to being over-ambitious. Fortunately, Bikerdude has helped me overcome most of them. Maybe. Not as holy writ, but as design tips, sure.