One page. Six sections, in this order:

Problem. One paragraph. What’s actually broken, for whom, right now — not the version of the problem that justifies the feature I already want to build.

Evidence. What I observed or was told that makes me believe the problem is real, not assumed. If this section is empty, the PRD doesn’t move forward yet.

Workflow today. How the person does this now, in their words, with the actual number of steps and handoffs — not the idealized version.

Workflow after. The same walkthrough, with the change in place. If this doesn’t visibly shorten or simplify the “today” version, the feature isn’t earning its complexity.

What I’m not building. The adjacent, tempting version of this that I’m deliberately leaving out, and why. This section prevents scope creep more than any other part of the doc.

How I’ll know it worked. A number or a behaviour, not a feeling. Something I can check in two weeks and be wrong about, if I was wrong.