I usually start from a concept, a central conceit or objective that will drive the story. Let's say my starting point is ... a vampire unicorn. (I am fairly certain I will not want to write about a vampire unicorn, but let's suppose I did.)
I need the story to go somewhere, and I don't write well unless I have an ending in mind. So for now I'll say the ending is a big confrontation with the vampire unicorn. It also needs to start somewhere, so we'll say someone is looking for the vampire unicorn. Why? Well, maybe the horn of the vampire unicorn is reputed to be a cure for the curse of vampirism, and the seeker is trying to cure someone who's a vampire.
Now I have a protagonist and at least one supporting character. But who's the protagonist? The person seeking the vampire unicorn? Or the vampire he's trying to cure? And is the vampire unicorn a character, or a mysterious figure of legend unseen until the final culmination of the story?
I keep asking myself these kinds of questions until interesting characters emerge. The unicorn-seeker is now a self-styled hero of derring-do and nobility. But he's trying to alter another person against that person's will (the vampire). If the vampire is not really all that bad, and is not doing any harm to others, that suggests that our unicorn-seeker is misguided -- either a protagonist with a serious flaw or an antagonist whose goals are fundamentally opposed to our protagonist. I'm liking the latter option, so I start asking myself whether the vampire is an interesting protagonist, or if there is an as-yet undiscovered character waiting in the wings to take on our lead role.
Once I have an intriguing cast of characters -- one or more protagonists, a really good antagonist, and some supporting foils if I'm feeling in the mood for juggling plotlines -- I pick a Point A to go with my Point VU (Vampire Unicorn), and then I work up a rough outline that takes me from one to the other. In this case, I declare that my Point A will be the moment in which the unicorn-seeker boldly sets forth on his trek, because this will allow him to make a nobly heroic speech and will allow me to challenge myself to craft a scene in which a seemingly upright and admirable figure is strikingly drawn, but that also leaves the readers just ever-so-slightly unsettled -- hopefully for reasons they can't quite put a finger on.
Point B switches to our vampire, whom the hero has portrayed in his speech as the suffering victim of a curse, and whose salvation he intends to arrange by dint of mighty deeds. But upon meeting our vampire, we find she's a relatively content but perhaps slightly gloomy individual whose only curse is the persistent attentions of the unicorn-seeker, whom she regards with aggravation. A supporting character informs her of the impending quest, and she determines to oppose its completion.
The story is looking itinerant at this point, so I start throwing out some locales and events that our contending duo may pass through, or people they may meet, as they separately try to discover the vampire unicorn's location and reach it first. These may end up as chapter titles, or may be discarded or replaced as I work on the outline in more detail:
The Sewers of Sarville
Incidents Under a Library
Two Old Fishwives
On the Road to Wougliefe
A Dragon's Outhouse
The Fallen Mountain
Into the Earth's Bones
The Vale of Starlight
I may dive in and try to write an opening scene or two with little more to go on than what I've spelled out here, or I may spend days or weeks making notes about the outline, trying to figure out where the paths of our protagonist and antagonist will cross, and how their relationship will develop in the process. In any case, by this point I will have started a composition notebook to keep my thoughts in, where I will repeatedly outline and re-outline as work on the story progresses.
When I have a solid enough feel for the characters and the general path of the story, I start writing, and try to hold myself to a strict-but-not-too-strict routine. I make myself write on the project almost every day, allowing myself one day off a week, perhaps two if family matters intervene. I may only write a paragraph in a sitting, but I keep at it day in and day out, pushing forward.
I follow a few simple rules as I write.
First, I try my best to avoid rereading any prose unless I absolutely have to. Unless I need to remind myself of some detail from chapter one, I don't go back to it until the book is done. Rereading makes me tempted to rewrite things, and for me it's pointless to rewrite while the first draft is still in progress. I have to maintain forward inertia and be prepared to edit ruthlessly in subsequent drafts.
Second, whenever I encounter a scene (or even a whole chapter) that is slow and tedious to write because it just doesn't grab my attention, I throw it out and look for a way to rework that part of the outline. If the story isn't keeping me interested, why would I expect it to interest a reader?
Finally, I absolutely refuse to allow my characters to be stupid just because it advances the plot or helps move the story in a direction I want it to go. If I have designed a character as a stupid individual, then by all means that stupidity can drive the plot in any number of ways. But if I want the reader to admire a character, I never permit the character to overlook something obvious or make a foolish mis-step. My characters are not infallible; they err all the time. But those errors always reveal something about their personalities or help to develop the story's theme. They're never shortcuts that allow me to avoid working harder on my plot.
One more rule that I have found very useful is this: when I'm done with the first draft, I put the book aside and don't even look at it for several weeks or months. Someone told me that Stephen King does this, and I find that it has two terrific effects. First, it gives me the closest experience possible to the experience of a reader who's never read the book at all. And second, by dulling my familiarity with the prose, it allows me to more objectively judge what works and what doesn't, so that I have a clearer view of what to fix as I approach the next draft.
After that "hide it in a drawer" period, I adjust the manuscript into paperback format and print it through an online print-on-demand site, making copies for myself and a few trusted friends on whom I rely for initial feedback. Then I reread the book start to finish, just for the reading experience, making notes only when I see something that I'm worried I'll forget about. Then I compare my reactions with those of my first readers, usually over a lunch or dinner where we can chat at ease and I can make more notes.
My revision process then consists of one major pass and one or two editing passes. In the major pass, I'll add, delete, or rewrite whole scenes, perhaps to the tune of 10 or even 20% of the original prose. The editing passes are mostly about tweaking the language, finding things that I missed or errors I introduced in the major pass, and making sure my continuity is as seamless as possible.
Once I'm happy with the final product, the hardest work starts, which is trying to sell the book to a publisher. I love the mental puzzle of outlining a story, and the challenge of wrestling the outline into prose form, but for some reason, I absolutely hate to write summaries, query letters, and cover letters. If I had my way, all the book jackets of my novels would have the same back-cover text: "Just read the first page, already. If you think it's interesting and well-written, you'll probably like the rest too."
In a nutshell, then, that's how I do what I do. I've written four books this way now, after stumbling around to develop the process on a fifth that won't likely see publication anytime soon. It's been working pretty well for me, and from what my readers tell me, the results are working for them too.
Now I just have to fight off the temptation to set aside my current book to start writing about that vampire unicorn ...