Skip to content

PRD Template: write requirements with a clear reason

This Product Requirements Document Template connects a product decision to observable behavior. Use the blank structure for a real initiative, then adapt its detail to the uncertainty and consequences of the change.

Educational resource · PM Harness Concept & Knowledge Hub

When to use this PRD template

Use a PRD when several people need a shared explanation of what should change and why. It is especially useful before a handoff to an engineer or coding agent, when a conversation alone would leave the scope ambiguous.

For a small reversible change, keep the document short. For a change with permissions, important dependencies, or difficult rollback, expand the relevant constraints and failure scenarios. Do not add sections merely to make the document look substantial.

A PRD records an accepted product choice. If you cannot explain the problem or compare alternatives, first return to the research and decision stages of PM Harness.

Copy or download the Markdown template

The fields below are intentional prompts to fill in, not approved requirements. Mark unknowns explicitly instead of guessing.

Download the PRD template as Markdown

# Product Requirements Document

## Document control
- Initiative / version:
- Owner / reviewers:
- Status: Draft | In review | Approved | Superseded
- Approved decision reference:

## Problem and users
- Who experiences the problem?
- What happens today, and why does it matter?
- Which user group is excluded from this version?

## Evidence and uncertainty
| Source ID | Observation | Date | Limitations |
| --- | --- | --- | --- |
| E1 | | | |
- Assumptions still requiring validation:
- Contradictory evidence:

## Goal and measurement
- Intended user outcome:
- Current baseline, or plan to establish it:
- Proposed success criterion and why it is meaningful:
- Guardrail / adverse outcome to watch:
- Measurement owner and review date:

## Decision and alternatives
- Accepted approach and rationale:
- Alternatives considered, including doing nothing:
- Conditions that would reopen the decision:

## Scope and non-goals
- Included behaviors:
- Explicit exclusions:
- Dependencies and constraints:

## Requirements
| ID | Required behavior | Rationale / source | Acceptance reference |
| --- | --- | --- | --- |
| R1 | | | A1 |

## Acceptance scenarios
- A1 Given [starting condition], when [action], then [observable result].
- Permission / unauthorized scenario:
- Empty or invalid input scenario:
- Failure / retry scenario:
- Relevant accessibility behavior:

## Rollout and verification
- Initial audience and release conditions:
- Checks required before release:
- Product outcome review after release:
- Pause / rollback conditions and owner:

## Open questions
| Question | Owner | Blocking? | Resolution needed by |
| --- | --- | --- | --- |

## Approval and changes
- Accepted scope and version:
- Approver and date:
- Subsequent changes, rationale, and linked decision:

Filled example: a filtered activity export

Illustrative feature: an operations analyst needs to review a small set of workspace activity outside the application. The following is a fictional example, not a shipped feature or a claim about customer behavior.

  • Problem: an analyst can filter activity on screen but cannot give a reviewer the same filtered record.
  • Goal: let an authorized analyst export the currently visible filter selection without exposing records outside their access.
  • Decision: provide a manual CSV export. Defer scheduled exports and public sharing links.
  • R1: an export contains only records the user is permitted to view and that match the selected filters at export time.
  • A1: given a workspace with two activity categories, when an authorized analyst exports with one category selected, the file contains only authorized records in that category.
  • Failure case: if export generation fails, show a recoverable error and do not provide an incomplete file as a successful export.
  • Open question: what volume can be handled synchronously? Engineering must establish this before an interface promises immediate completion.

A proposed outcome review would observe whether analysts can complete the intended review task. Successful file creation is a technical check, not proof that the review workflow improved.

Adapt the template without losing its purpose

For an early experiment

Emphasize the hypothesis, audience, smallest useful behavior, and learning decision. Do not describe a complete future platform if only one assumption is being tested.

For an existing product change

Include current behavior, compatibility expectations, affected users, and how support or operations will handle the transition. Identify which earlier decision is being revised.

For a coding agent

Add a reviewed technical handoff with relevant system boundaries, test scenarios, and escalation rules. Keep those details connected to the product requirements; use the agent specification guide rather than turning every PRD into an architecture document.

Review the document before calling it ready

Check that every requirement has a reason and that every acceptance check points to required behavior. A goal with no measurement plan cannot be evaluated; a metric with no user purpose can reward the wrong outcome.

  • Can another person identify both included behavior and non-goals?
  • Are evidence and assumptions distinguishable?
  • Are permission, invalid-input, and failure cases addressed where relevant?
  • Does each blocking question have an owner?
  • Is the approved version identifiable after later edits?

Do not use a word count or section count as a readiness test. Keep a decision record separate when the rationale is substantial, and link it from the document. A product decision log explains that relationship.

For an independent overview of PRD fundamentals, consult Atlassian's PRD guidance. This downloadable template and its example are original to this educational site.