PRD · Product management · Templates
How to write a PRD: what goes in it, with a filled-in example
Last updated By T. Krause
A product requirements document says what a feature should do and why, before anyone designs or builds it. Here are the sections a PRD needs, how a small team keeps one short, a complete example for an invented booking app and how to keep it current during the build.
A PRD, or product requirements document, describes one feature before anyone designs or builds it. It says what problem the feature solves, who has that problem, what the feature must do, what it won't do and how you'll know it worked.
To write one, start with the problem and the evidence for it. Name the users, write two or three goals and the things you are deliberately leaving out. Then list the requirements sorted into must, should and could have, sketch the user flow, set the success measures and write down every question you can't answer yet. Put an owner and a status at the top, share it for comments and keep it updated until the feature ships.
Below, each section gets a short explanation, followed by a complete PRD for an invented feature at an invented four-person company. If you'd rather start from a page with the sections already in place, use our PRD template.
What a PRD is, and when a small team needs one
A PRD covers the what and the why of one feature. It leaves the how to the people building it. "Members can join a waiting list for a full class" belongs in a PRD. The database table that holds the list belongs in the design doc that engineering writes afterwards.
The person who owns the product decision writes it. In a larger company that is a product manager. In a team of four it is often the founder, or whoever will answer questions about the feature for the next two months. Designers, engineers and support read it, comment on it and help fix it before work starts.
You need one when a feature takes more than a couple of weeks, when more than two people work on it, or when someone outside the team has to agree to it. A change one developer can finish in two days doesn't need a PRD. A ticket with three lines of context will do.
PRD, design doc and one-pager
These three get confused because they all come before the build.
- A project brief or one-pager asks whether something is worth doing at all. It has the goal, the rough scope, the owner and the cost, and it is short enough for a manager or client to approve in one sitting. Our project brief template has that shape, and the guide on how to write a project brief walks through a filled-in one.
- A PRD comes once the answer is yes. It turns the goal into requirements that design and engineering can work from.
- A design doc, or technical spec, explains how engineering will build it, which alternatives they rejected and how they will roll it out. See the design doc template, and our guide on how to write a design doc for a filled-in one.
Small teams often merge the brief and the PRD into one page, and that's fine. Keep the design doc separate, though, because the two have different readers and change for different reasons.
Why many PRDs go unread
Marty Cagan of the Silicon Valley Product Group wrote the classic complaint in his 2006 essay "Revisiting the Product Spec". He said most specs "take too long to write" and "are seldom read", and that the spec's mere existence can tell management and the team that everything is going fine when it isn't. His answer was to make a high-fidelity prototype the main part of the spec, because you can put a prototype in front of real users and a paper document you can't.
For a small team the lesson is to keep the PRD short and let screens or a clickable prototype carry the interface detail. The PRD then holds what a prototype can't show: the problem, the goals, the things you're leaving out, the rules behind each screen and the questions still open.
What goes in a PRD
These are the sections in our PRD template, in the order it uses them. Other templates use different names for some of them, but the content is much the same.
- Status line. One line at the top with the status, the owner and the reviewers. Readers should see at once whether the document is a draft, in review or agreed, and who to ask.
- Problem. What goes wrong today, for whom, and the evidence: support emails, interview notes, things you watched people do. Link the sources. If you can't describe the problem without describing your solution, you haven't found it yet.
- Users. The groups who have the problem, in a line each. If the feature has two kinds of user, say a member and an admin, name both, because each needs different things.
- Goals. Two or three outcomes in plain words. "A freed place gets filled without the studio lifting a finger" is a goal. "Build a waiting list" is a task.
- Non-goals. What this feature will not do, even though someone will ask. Writing it down now saves arguing about it in week three.
- Requirements. What the feature must do, sorted into must have, should have and could have. Number the musts so reviewers can say "I disagree with 3" instead of quoting it. Write each one as something a tester could check.
- User flow. The main path through the feature, in numbered steps from the user's point of view. Link the designs here once they exist.
- Success metrics. How you'll know the goals were met, with a number and a time frame. Decide them now, because after launch every result looks like success.
- Open questions. Everything you don't know yet, each with a proposed answer if you have one and the name of whoever will settle it.
- Timeline. The dates that matter for this document: when review closes, when the build starts, when the first users get it.
Leave out what nobody will read. A small team doesn't need a glossary, a revision table with version numbers or a page of background on the market. One thing the template lacks and we'd add is a short Changes list at the bottom. The section on keeping the PRD current explains why.
How to write a PRD in five steps
1. Write the problem before you touch the solution
Collect the evidence first. Read the support emails, the call notes and the requests in your feedback board, and write the problem in three or four sentences that someone outside the team would understand. Then show it to whoever talks to customers most and ask whether it matches what they hear.
2. Settle the goals and non-goals with the decision maker
Before you write a single requirement, agree on the goals and the non-goals with whoever can say yes to the feature. If these two lists aren't agreed, every requirement gets argued over one by one.
3. List the requirements, then cut
Write every requirement you can think of, then sort them. A must have is something without which you would rather not ship. Expect the first list of musts to be too long, and move anything that doesn't serve a goal down to should or could.
4. Write the flow and the questions
Walk through the feature as the user would, step by step. Every place where you write "it depends" becomes an open question with a name next to it.
5. Share it with a deadline
Share the PRD for comments, name the reviewers and set a date when review closes. Answer each comment in the text and resolve the thread, so the page always shows the current answer instead of a debate.
A filled-in example PRD: waiting lists for full classes
The company, the people, the dates and the numbers below are invented. Fernhill is a four-person company that makes a class-booking app for small yoga and pilates studios. Priya Nair is the founder and owns the product. Tom Ellis writes the code, Joanna Kowalski does design and Ben Adeyemi answers support. Priya wrote the first draft on Wednesday 14 October 2026 and revised it a week later.
PRD: waiting lists for full classes
Status: in review until Wednesday 28 October 2026. Owner: Priya Nair. Reviewers: Tom Ellis, engineering. Joanna Kowalski, design. Ben Adeyemi, support. Last updated: 21 October 2026.
Problem
When a class is full, members can't book it and the app gives them nothing else to do. When someone cancels, the place sits empty unless a member happens to check the app at the right moment. Studio owners email support to ask for a way to keep a list of people who want in. Ben gets one of those emails most weeks. Today some owners keep the list on paper at the front desk and text people when a place opens.
Users
- Members who book classes at a studio through the app
- Studio owners and front-desk staff who manage the timetable
Goals
- A place freed by a cancellation goes to someone who wants it, without the studio doing anything
- Studio owners stop keeping waiting lists by hand
Non-goals
- Waiting lists for multi-week courses, which book differently
- Text message notifications
- Charging a fee to join a waiting list
Requirements
Must have
- A member can join the waiting list for a full class, and leave it again.
- When a booked member cancels, the first person on the list gets an email and a push notification offering the place.
- The offer holds for 15 minutes. If it isn't accepted, it goes to the next person on the list.
- Accepting the offer books the class in one tap.
- The studio owner sees the waiting list on the class page and can remove someone from it.
Should have
- Members see their position on the list
- Studios set a maximum length for each class's list
Could have
- Members choose to be booked straight away when a place opens, without the offer step
User flow
- A member opens a full class and taps Join waiting list.
- The app confirms they're on the list.
- Someone cancels. The first member on the list gets an email and a notification with an Accept button.
- The member taps Accept and the class appears in their bookings.
- If 15 minutes pass without an answer, the offer moves to the next member on the list.
Success metrics
Measured over the four weeks after release, from 7 December 2026 to 4 January 2027:
- At least half of the places freed by cancellations are filled from a waiting list
- Support receives fewer than five emails a month asking for waiting lists
Open questions
- Studios that charge per class: does the member pay when they accept the offer? Proposal: yes, with the card already saved. Tom to check with the payment provider by 28 October.
- Should offers go out in the last hour before class? Proposal: no, the place opens for normal booking instead. Joanna to ask two studios by 28 October.
Timeline
- Review closes: Wednesday 28 October 2026
- Build starts: Monday 2 November 2026
- Beta with five studios: Monday 23 November 2026
- Release to all studios: Monday 7 December 2026
Changes
- 21 October 2026: offer window cut from 30 minutes to 15. Ben pointed out that evening classes often lose and refill a place within the hour, and a 30-minute hold would leave it empty for too long. Priya agreed and logged the decision.
- 14 October 2026: first draft.
A few things to notice. The problem describes what goes wrong for members and for studio owners, with evidence Ben can point to, before any requirement appears. The non-goals answer three requests before anyone makes them. Every must have is something Tom can build and Ben can test. The two open questions each have a proposed answer, a name and the same deadline as the review, so the review can't close with them still open. And the Changes list shows why requirement 3 says 15 minutes, which saves Tom asking in February.
How to keep the PRD current while the feature is built
A PRD that stops changing the day the build starts is soon wrong. Engineers find a case nobody thought of, or a requirement turns out to cost far more than the estimate. If the PRD doesn't record that, people go back to asking in chat and the document turns into the false comfort Cagan warned about.
- One owner keeps it. The person who wrote it stays responsible for it until the feature ships. Anyone can point out that it's wrong. The owner decides how the text changes.
- Update it the day a decision is made. When the team changes scope in a meeting or a chat thread, the owner changes the requirement that day and adds a dated line to the Changes list with the reason.
- Log the bigger decisions. A decision that matters beyond this feature, such as a rule that will apply to every future booking feature, also goes in the team's decision log with a link to the PRD. Our decision log guide shows what an entry needs. If the decision came out of a meeting, link the meeting minutes as the source.
- Answer open questions in place. When a question is settled, write the answer under it, note it in Changes and move whatever it changes into the requirements.
- Close it at release. Once the feature ships, set the status to Shipped and add the release date. When you check the success metrics, add the results under them. The PRD then tells the next person what was built and whether it worked. Customers get the short version in the release notes.
Writing a PRD in Prompt Notebook
Prompt Notebook is ours, so read this part with that in mind. The product requirements document template opens as a page with a status callout at the top naming the owner and reviewers. Below it are Problem, Users, Goals, Non-goals, Requirements split into Must have, Should have and Could have, User flow, Success metrics, Open questions and Timeline. Open questions sit in toggles, with the proposed answer inside. The sample is a PRD for recurring invoices at an invented invoicing app, so you see a finished one before you replace it. The template has no Changes section. Add one at the bottom if you follow the advice above.
Write the problem and goals yourself, then use Ask AI to continue with a first pass at the requirements. Expect to cut that list, as in step 3. Link the support tickets or interview notes in the problem section with @, so each source page lists the PRD under Linked from. Share the page with comment access and tag the reviewers in the status callout. Page history takes a snapshot every 10 minutes while the page is being edited, so you can see what changed since the review and restore an earlier version.
Frequently asked questions
How long should a PRD be?
Long enough that design and engineering can start without a meeting to explain it. For one feature at a small team that is usually one to three screens. If it runs much longer, check whether it covers two features, or whether you've written the how that belongs in the design doc.
Who writes the PRD?
The person who owns the product decision for the feature, usually a product manager. In small teams that is often the founder or a lead. Engineers, designers and support review it, and the owner decides what changes.
What is the difference between a PRD and a spec?
A PRD says what the feature should do and why. A technical spec or design doc says how engineering will build it. Some teams use "spec" for both, so ask which one someone means before you start writing.
Do agile teams still write PRDs?
They can, and a short one helps. User stories and tickets describe pieces of work, but they rarely hold the problem, the goals and the non-goals in one place. A short PRD that the tickets link to gives everyone the reason behind the stories.