I write the last chapter first. Before I draft a page of the story, I know who did it, how, and why, and I know the one physical thing the crime left behind. That last piece matters most, because it is what my detective will find. Everything else in the book is built to put it in her path.
Backward from the trace
A real failure leaves a real mark. A part that was forced shows the force. A tool used wrong leaves a signature the right tool would not. I spent twenty years reading those marks after the fact, and a story lets me plant one on purpose, then arrange for someone to leave it.
So I start at the trace and work backward. The crime left this mark, the mark came from this act, and the act was this person's, for this reason. Told forward, that chain is the story. I build it in the other direction.
An ordinary detail
The mark has to be on the page early, and it has to look like nothing much. This is the part that takes the work. The thing that solves the case sits in an early chapter looking like an ordinary detail of an ordinary day, a receipt kept or a chair moved from one room to the next. A reader goes past it because it looks like part of the day and nothing more. It has to be genuinely dull the first time through and genuinely decisive the second, and making one object do both is most of the job.
A page for every clue
I keep a clue ledger, a page for each clue. On the page I write four things. What the object is. Where in the book it appears. What it truly means. And what else could plausibly explain it, some ordinary account a reasonable reader would reach for first. I use a bound ledger and not index cards, because cards get shuffled and lost, and a ledger stays in order and stays whole, and order is the entire point of the exercise.
Walking backward keeps me straight
Working backward is what keeps the ending from cheating. A writer who works forward and reaches the last chapter needing one more fact is tempted to reach back and slip it into an earlier scene, so that the solution rests on something the reader was shown too late and dressed up as if it had been there all along. Building out from the solution closes that door. Everything the ending needs is on the ledger before I write the middle, so there is nothing left to smuggle in. If a fact is not in an early chapter, it cannot be in the last one.
This is the promise a fair-play mystery makes, that the reader could have gotten there first, and the only way I know to keep it is to know the end before I write the start. Fixed to Death, my first novel, comes out on August 4, and it was built this way from the last page forward. If you read it twice, the second time is the one where you catch me.