Product requirements document (PRD)
Also called a product requirement document.
A product requirements document (PRD) describes what a product or feature must do, who it serves, and how the team will judge success. It gives product, design and engineering a shared scope before development starts, without prescribing every implementation detail.
Key takeaways
- A PRD connects a user problem to specific requirements, acceptance criteria and a success measure.
- State what is out of scope. Otherwise, a small feature can quietly become a much larger project.
- Acceptance criteria test whether the feature works. Success metrics test whether it solves the customer's problem.
- For a startup, the useful PRD is the one that prevents wasted development, not the one with the most pages.
How a PRD works
The product owner drafts the requirements, then reviews them with the people who will design, build and test the feature. Unresolved questions should remain visible rather than become hidden assumptions.
A useful PRD covers:
| Section | What it needs to answer |
|---|---|
| Problem and user | Who has the problem, and what evidence shows it matters? |
| Goal and success measure | What should improve after release, and how will you measure it? |
| Scope and exclusions | What is included now, and what is deliberately deferred? |
| Requirements | What must users and the system be able to do? |
| Acceptance criteria | What observable result means each requirement passes? |
| Dependencies and decisions | What must happen first, who decides, and what remains unresolved? |
The document should change when the team learns something material. Record scope changes and their effect on delivery estimates; silently rewriting requirements makes it hard to understand why a deadline moved.
A PRD does not replace conversations with customers or engineers. It records the decisions those conversations produce.
Cutting an MVP from 18 to 10 engineer-days
A startup's five pilot agencies want clients to approve design work without long email threads. The founder initially requests approval links, notifications, custom branding and reporting.
The PRD narrows the goal: at least four of the five agencies should complete an approval without staff help during the first week.
The following estimates are illustrative, not industry benchmarks. They include development and testing.
| Requirement | Estimated engineer-days | Initial scope | MVP scope |
|---|---|---|---|
| Create and share an approval link | 4 | Included | Included |
| Record a decision and timestamp | 4 | Included | Included |
| Notify the project owner | 2 | Included | Included |
| Customize client-facing branding | 3 | Included | Deferred |
| Build an approval analytics dashboard | 5 | Included | Deferred |
| Total | 18 days | 10 days |
At an illustrative contractor rate of $700 per engineer-day:
- Initial development budget: 18 × $700 = $12,600.
- Narrowed MVP budget: 10 × $700 = $7,000.
- Difference: 8 engineer-days and $5,600.
This is a scope reduction, not a productivity gain. Branding and analytics remain possible later.
One acceptance criterion could require that submitting an approval saves the decision and timestamp, displays both to the project owner, and does not create duplicate records if the client retries. That is testable. "Make approvals seamless" is not.
The founder gets a cheaper test of the customer need before committing to the rest.
What a published PRD template includes
Atlassian's Product requirements template includes objectives, success metrics, assumptions, requirements, user interaction and design, open questions, and items outside scope.
The important pattern is that requirements sit alongside their purpose and boundaries. A list of requested features alone leaves the team to guess which customer outcome matters and what can wait.
Use the template as a starting point, not a requirement to fill every section. A small change may need only a short document; a feature with several dependencies needs more explanation.
What to question before development starts
PRD versus roadmap versus technical specification
| Document | Main question | Typical contents |
|---|---|---|
| PRD | What must this product or feature do, and why? | User needs, scope, requirements, acceptance criteria |
| Product roadmap | What outcomes or initiatives come next? | Priorities, sequence, planning horizons |
| Technical specification | How will we build it? | Architecture, interfaces, data models, implementation decisions |
Teams sometimes combine these documents. The distinction matters more than the file name.
Questions people ask
Who writes a PRD?
A product manager often owns it, with input from engineering, design and other affected teams. At an early-stage startup, a founder may write it. The people responsible for delivery should review feasibility and acceptance criteria before work begins.
Does an MVP need a PRD?
It needs shared requirements, not necessarily a formal document. Write enough to identify the hypothesis, minimum scope, exclusions and pass conditions. Skip elaborate documentation that does not change a decision.
Can a PRD be a PDF?
Yes. A PDF works as a dated snapshot for review or approval. Keep an editable source and identify the version, owner and date so readers know whether they have the current requirements.
