You are a senior product manager at Learnify, an online learning platform with over two million registered users. Learnify offers courses across technology, design, business, and creative skills. The catalogue currently contains 4,800 courses across 60 categories.
The business has a serious problem. Catalogue growth has outpaced the platform's ability to help users navigate it. Users who browse the catalogue convert to enrollment at only 11%. Exit surveys and user research point consistently to the same theme: "I didn't know where to start" and "I couldn't find courses relevant to where I am in my career."
Meanwhile, the data team has surfaced something encouraging. Users who enroll in a course within their first two sessions have a 68% course completion rate. Users who take more than four sessions to enroll have a 19% completion rate. Early course-to-user fit predicts almost everything that matters downstream: completion, re-enrollment, subscription renewal.
You have been asked to design and spec a personalised course recommendation system that helps users find courses relevant to their skill level, career goals, and learning history, across the home feed, the course detail page, and a weekly email digest.
This is a large initiative. It involves the data science team, the engineering team, the design team, the email platform team, and has implications for legal due to personalisation data processing. It will take approximately one quarter to build.
Using the template below, write a complete PRD for this initiative. This template is more detailed than Exercise One because the feature is more complex. Every section will require more thinking, more specificity, and more deliberate choices about what to include and what to defer.
This exercise will be uncomfortable in places. That is the point.
Document Title: Author: Status: Version: Last Updated: Stakeholders:
Ask yourself: who holds approval authority for a feature of this complex? Who are the contributors whose input must shape the document before it is finalised? Who needs to be informed without being directly involved?
Overview
Write three to four paragraphs. Set the strategic context. Explain why this initiative matters now. Describe what is being built at a high level. State clearly what this version covers and what it does not.
Problem Statement
This is a complex problem with multiple contributing factors. Write a problem statement that captures the core user experience failure, the business impact, and the evidence for both.
Resist the temptation to name the solution.
We have observed that __________________
We know this because __________________
The impact of leaving this problem unsolved is __________________
Goals
Write four goals. Given the complexity of this initiative, your goals should address both the immediate user experience improvement and the longer-term business outcome. Each goal must describe an outcome that is measurable.
This feature touches multiple surfaces and multiple user types. Define metrics for each dimension of success. Include primary, secondary, and guardrail metrics. Pay particular attention to guardrail metrics: what must not get worse as a result of this feature?
Primary Metric:
Secondary Metrics:
Guardrail Metrics:
Evaluation Timeframe:
Ask yourself: what is the earliest signal that would tell you the feature is working? What is the lagging indicator that would tell you it is truly succeeding? Are you measuring both?
User Personas
Define three distinct user personas. Each persona should justify at least one requirement that the other personas do not. If removing a persona would not change any requirement, the persona is not doing its job.
Persona One: Name, context, goals, frustrations, risk if the feature fails them
Persona Two: Name, context, goals, frustrations, risk if the feature fails them
Persona Three: Name, context, goals, frustrations, risk if the feature fails them
User Stories
Write five user stories, at least one for each persona. Make each story specific enough that it points toward a concrete requirement. A user story that could apply to any product is not specific enough.
Functional Requirements
Organise requirements by feature area: Recommendation Engine, Recommendation Surfaces, Feedback Mechanisms, and Onboarding. Within each area, use Must Have, Should Have, Nice to Have prioritisation. Write each requirement as a specific, testable behaviour.
Recommendation Engine
Must Have:
Should Have:
Recommendation Surfaces
The feature must appear in at least three surfaces. For each surface, specify the requirements separately.
Home Feed:
Course Detail Page:
Weekly Email Digest:
Feedback Mechanisms
Must Have:
Onboarding
How does the system handle new users with no learning history? Specify the requirements.
Must Have:
Ask yourself: what happens when a user has no history? What happens when a user's interests change? What happens when the system recommends something and the user ignores it repeatedly? Each of these edge cases belongs in your requirements.
Non-Functional Requirements
A personalised recommendation system has significant non-functional requirements. Address all of the following: performance, scalability, explainability, fairness, privacy, and accessibility. For each one, write a specific, testable requirement.
Out of Scope
This is a large initiative with many adjacent possibilities. Name at least five things that are explicitly not in scope for this version. Be specific enough that a stakeholder reading this section cannot claim they assumed otherwise.
Dependencies
A feature of this complex will have multiple dependencies across teams and systems. Name each dependency, identify which team owns it, and state what must be true before development can begin or before launch can happen.
Dependency | Owner | Required By |
Risks
Name at least four risks. For a personalised recommendation system, think beyond technical risks to include risks related to user experience, creator ecosystem health, and data ethics. For each risk, be honest about likelihood and impact.
Ask yourself: what could go right in the metrics but wrong for users? What could go right for most users but wrong for a specific group? What could go wrong in the first 30 days that you would not see in your headline metrics?
Risk | Likelihood | Impact | Mitigation |
Open Questions
A first draft of this PRD will have more open questions than a simple feature PRD. Name at least five. Assign each one an owner and a deadline. Do not carry open questions into development without a plan to resolve them.
Question | Owner | Deadline |
Reflection Questions
When you have finished your draft, sit with these questions before you consider it complete.