Here's my two cents on this, since wrote a Master's Thesis on marketing films [humblebrag achieved!
], so I feel that I know a little something about trailers.
Give potential customers as much information as possible so they can make up their mind. That's the most important thing. Have a clear idea what you want to communicate with your trailer, who your audience is and what they want consider important for them. A casual audience may want a game that's fun, can be played in short sessions and doesn't force them into micro transaction down the road, for example; and RPG audience is interested in individualization options and depth in both gameplay and story, etc. In any case, research your intended audience first from a marketing perspective. I know, that sounds pretty boring but your trailer will be *the* deciding factor for giving your game a chance or not.
Don't mislead. If you have a puzzle game with puzzles that you consider pretty hard, for example, don't gloss over it and try to sell it as an easy little game for everyone, instead own it and tell people right away that the puzzles are hard but rewarding. If you have an endless runner with some light RPG elements, don't sell it as an RPG with endless-runner elements. People looking for a decent RPG to play will be pissed because the game is lacking depth, while more casual players who are looking just for something to pass the time at the airport will stay away because they might feel it's too involved for what they want. Wither way is bad, people feel cheated and will rate your game poorly regardless of its quality.
Effectively, a game trailer is there to answer questions, so first you have to find the questions your audience might ask, what their tastes are. Start broadly. What's the genre? What's interesting about your game? What separates it from other, similar games? Answering these questions visually means helping your potential customers decide whether your game is a fit for their tastes. Here's a list of questions and answers in written form. Once you have a bunch of these, order them in a meaningful way (=from broad to specific) and find a visual answer for each, i.e. gameplay footage.
- Q: "What kind of game is it?" Answer: "A match-three puzzle game for mobile"
- Q: "What's different about this one?" A: "It offers light RPG elements and customization"
- Q: "How much depth is there?" A: "It's easy to get into and there are 20+ levels"
- Q: "How long are the play sessions?" A: "A level takes from 10 seconds to 5 minutes"
- Q: "Can I play with friends?" A: "Your Facebook friends can send you gifts but that's entirely optional"
- Q: "What does it cost?" A: "Free to play"
- Q: "Where's the catch then?" A: "You can pay for boosters that help you level faster and cosmetic items"
If you can't show something, it's perfectly fine to use a bit of text, e.g
- "20+ levels!" (video in background shows some of that variety)
- "Light RPG Elements" (video shows snippets of level-up animation, and distribution of points between "Strength" and "Mana")
Structure:
I'd go for a couple of seconds of gameplay footage, then a text plate. That's pretty standard but you won't be too original in your trailer, instead show people what they expect from a trailer. So text plates between some gameplay work especially well on mobile because a majority of people have their sound switched off when browsing the Play store when riding the train to work, for example, so focus in the visuals but also have some music and sound, but it's not that important. These text titles should be short, both in content and length, and, again: Should help in answering questions.
Length:
Trailers usually range from 1 to 3 minutes, but especially on mobile I would aim for 20-30 seconds. Show the most interesting part as early as possible to keep potential players watching, but don't throw it at them without some context of what your game is about. The first impression is what sticks, so if you have a memorable character, an interesting look / mechanic that's easy to grasp in 2 seconds, put that in front.
I hope that more theoretical replay was useful!