Requirements say what a system must do and how well it must do it, before anyone builds it. Most failed projects did not fail at coding: they built the wrong thing, or an unclear thing. Good requirements come from talking with the people who will use the system and writing down what you learned, precisely enough to test.
Key concepts
- Functional and non-functional: what the system does, and how fast, safe, available or usable it must be.
- Stakeholders: users, operators, buyers and regulators each want different things; find them all.
- Testable wording: "fast" is not a requirement; "95 percent of searches return within one second" is.
- Use cases and user stories: describe a goal from the user's side, with its normal and failing paths.
- Change and traceability: requirements shift, so keep track of why each one exists and what depends on it.
Vocabulary
- Acceptance criteria: the conditions that tell you a requirement is met.
- Scope creep: the slow growth of a project beyond what was agreed.
- Prototype: a rough version built to learn what is really needed.
- Specification: a written, agreed description of the behaviour of a system.
What to study, in order
- Interview someone about a task they do, and write what they need without proposing a solution.
- Turn five vague wishes into testable requirements with numbers and conditions.
- Write use cases for a small system, including what happens when things go wrong.
- Build a throwaway prototype and see which requirements it overturns.
- Review a real specification and list its ambiguities and missing cases.